CA2008370C - Compression of binary halftones - Google Patents
Compression of binary halftonesInfo
- Publication number
- CA2008370C CA2008370C CA002008370A CA2008370A CA2008370C CA 2008370 C CA2008370 C CA 2008370C CA 002008370 A CA002008370 A CA 002008370A CA 2008370 A CA2008370 A CA 2008370A CA 2008370 C CA2008370 C CA 2008370C
- Authority
- CA
- Canada
- Prior art keywords
- image
- lines
- line
- values
- given line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/41—Bandwidth or redundancy reduction
- H04N1/4105—Bandwidth or redundancy reduction for halftone screened pictures
Abstract
ABSTRACT OF THE INVENTION:
A system is described for reformatting halftone data for compression, wherein an original bilevel image is re-formatted to produce another bilevel image that allows vertical correlations to be recognized by the com-pression technique, thus improving compressibility dra-matically, with particular suitability for facsimile transmissions. In reformatting it is assumed that a se-lected halftone frequency H will satisfactorily describe an entire document, and each of successive sets of N
consecutive lines are concatenated to form respective single lines. The thus reformatted lines have a clearer halftone periodicity offering greater correlation and permit more efficient coding by well-known standard bilevel compression algorithms (e.g., CCITT G3 (MR) or G4 (MMR)). For an image with unknown pattern frequency, a technique for readily estimating the frequency for use in reformatting the image is described.
A system is described for reformatting halftone data for compression, wherein an original bilevel image is re-formatted to produce another bilevel image that allows vertical correlations to be recognized by the com-pression technique, thus improving compressibility dra-matically, with particular suitability for facsimile transmissions. In reformatting it is assumed that a se-lected halftone frequency H will satisfactorily describe an entire document, and each of successive sets of N
consecutive lines are concatenated to form respective single lines. The thus reformatted lines have a clearer halftone periodicity offering greater correlation and permit more efficient coding by well-known standard bilevel compression algorithms (e.g., CCITT G3 (MR) or G4 (MMR)). For an image with unknown pattern frequency, a technique for readily estimating the frequency for use in reformatting the image is described.
Description
-`" 20~83'Y~
COMPRESSION OF BINARY HALFTONES
BAC~GROUND OF THE INVENTION
.`' Field of the Invention This invention relates to digital ima8e processing S oethods and more particularly to methods for encoding and decoding image d~ta involving pattern frequencies such as binary halftones.
~"' Prior Ar~
The standard international data transmission CCITT Group 3 ~R and Group 4 MMR two-dimensional data compression schemes contain Modified Huffman code tables which were opti~ized for compressing text and line drsw~ngs. How~
~ver, blnsry halftone representations of continuous tone ~`
images have a very different distribution of run sizes and occurrence of vertical references from such drawings. As a result, the amount of data required to ~`~
represent these halftone images in "compressed" form ~ay ~0083~0 be greater thsn the amount of data required to represent the original image when ~hese compression schemes or ~ other currently used compression techniques are used.
This expansion can be limited to about 1.15 by using "uncompressed mode". However, by definition, this mode does not give any compression.
Some examples of prior systems for dealing wi~h the cc~pression of halftone image data are found in U.S.Pat No.4,144,547 to STOFFEL and U.S.Pat.No.4,559,563 to JOINER, which describe processes for coding mixed ttext and halftono) documents using ~;
multiple-predictor systems with sats of predefined pre-dictors. STOFFEL teaches the use of each availabla pre-dictor to predict esch unit of input data whereby tha best predictor is selected, its identity is encoded, and the unit of data is then encoded using that predictor.
The decoder gets an indication from the compre~sed data stream as to which predictor is to be used tb decode each u~it of dsta. JOINER uses the predictor which would have performed be~t on the previous unit of data to predict the pel values for the current unit of data ~so that the identity of the predictor to be used does not have to be transmitted). It will be seen thst these teachings are general approaches to the problem of efficiently , Y~989-001 - 2 -
COMPRESSION OF BINARY HALFTONES
BAC~GROUND OF THE INVENTION
.`' Field of the Invention This invention relates to digital ima8e processing S oethods and more particularly to methods for encoding and decoding image d~ta involving pattern frequencies such as binary halftones.
~"' Prior Ar~
The standard international data transmission CCITT Group 3 ~R and Group 4 MMR two-dimensional data compression schemes contain Modified Huffman code tables which were opti~ized for compressing text and line drsw~ngs. How~
~ver, blnsry halftone representations of continuous tone ~`
images have a very different distribution of run sizes and occurrence of vertical references from such drawings. As a result, the amount of data required to ~`~
represent these halftone images in "compressed" form ~ay ~0083~0 be greater thsn the amount of data required to represent the original image when ~hese compression schemes or ~ other currently used compression techniques are used.
This expansion can be limited to about 1.15 by using "uncompressed mode". However, by definition, this mode does not give any compression.
Some examples of prior systems for dealing wi~h the cc~pression of halftone image data are found in U.S.Pat No.4,144,547 to STOFFEL and U.S.Pat.No.4,559,563 to JOINER, which describe processes for coding mixed ttext and halftono) documents using ~;
multiple-predictor systems with sats of predefined pre-dictors. STOFFEL teaches the use of each availabla pre-dictor to predict esch unit of input data whereby tha best predictor is selected, its identity is encoded, and the unit of data is then encoded using that predictor.
The decoder gets an indication from the compre~sed data stream as to which predictor is to be used tb decode each u~it of dsta. JOINER uses the predictor which would have performed be~t on the previous unit of data to predict the pel values for the current unit of data ~so that the identity of the predictor to be used does not have to be transmitted). It will be seen thst these teachings are general approaches to the problem of efficiently , Y~989-001 - 2 -
2~8~70 Gompressing halftones and are not simple extensions of, but rather would be substituted for, the well~known ;
standard bilevel compression algorithms (CCITT G3/G4). ~`
.
U.S.Pat.No.4,355,306 to MITCHELL also describes a ge-neric bilevel image coder/decoder system. While this system psrforms well on halftones, it similarly would not operate as a simple extension of the standard bilevel compression algorithms or other commonly-usQd techniqnas.
. Other examples sre found in U.S.Pat.No.4,571,634 to `~
CANESCHI ET AL uhich describas a coder/decoder for bilevel and halftoned ims~es-in which the hslftones axe formed by having areas of strict alternation of black and white p~ls, thus restricting its applicability;
U.S.Pat.No.4,425,582 tQ KADAKIA ET AL and U.S.Pat.No.4,435,726 to LIAO which disclose an efficient hardware implementation for a specific predictor that ls designsd to work well on both text and halftone data ;
wherein the predictor and a corresponding te-predictor would be added to an encoder/decoder systam, such as ~ ;~
tha~ defined by the GCITT G3/G4 standards, to produce a version of the origlnal data altered by a predic~or scheme; and-U.S.Pat.No.4,193,096 to STOFFEL which shows Y09~9-001 - 3 _ 2[3~837~
a system that scans and halftones an image and then compresses the resulting image based on knowlsdge of the halftoning process used, so that compression is depend-ent on tha halftoning process generating the imsge dat~
and thus this teaching i~ inapplicable in a system where all that is supplied to the coder is the bilevel image,-ant the coder has no control over the ha}ftoning proc~
ess.
i Gonsequently, it is desirable, and an object of the present invention, to provide a simple and versatile technique that facilitates the ef~icient compression of binary halftone data representing continuous tone images by known and widely-uset compression processes. -.
SUMMARY OF THE INVENTION
The system and method of the present invention is di-rected to taking a binary halftoned image with a given halftone pattern frequency and reformatting it to obtain image data which compresses better than the original image data using the CCITT standard Group 3 or Group 4 two-dimensional compression techniques or other cur-rently popular data compression processes. Accordingly, each H lines of the original image srs concatenated into ~, - . -~
20(~8370 the form of a single image line`for purposes of com-pression, where H is the halftone pattern frequency. The resulting reformatted Lmage tata may then be compressed using one of the CCITT algorith~s, or with any othar algorithm which makes use of correlations between fea-tures on a current line and those on the immediately preceding line, in an efficient manner.
fur~her feature of the present invention is the de-termining, for an ~mags with unknown pattern frequency, of a good estimate of the frequency for use in refor-matting the image.
,. :
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a small piece of a typical halftoned image with a pattern fxequency of four, in both original and ~ ;~
raformattad form.
FIG. 2 contras~s the coding of a line of the original ~`
i~aga fragme~t using tha reference data from the (a) original ~nd (b) reformatted images.
FIG. 3 shows a method for estimating the halftone pat-tern frequency of an image.
Y098~-~01 - 5 -2~ i371:) DETAIT.~D DESCRIPTION OF A PREFERRED EMBO~IMENT OF THE
INVENTION
,~
The reformatting of ~ halftoned i~age in accordance with the present invention is illustrated in Figure 1. The ~ -~
upper image shows a small piece of a typical halftoned image. Esch square is one pel tpicture element). The small lines below and to the right of the image data indicate the demarcation between the pel rows and col- -u~ns. Ihe halftone pattern is formed Ln a 4x4 pel block.
The reformatted image is shown below with lines 1, 2,
standard bilevel compression algorithms (CCITT G3/G4). ~`
.
U.S.Pat.No.4,355,306 to MITCHELL also describes a ge-neric bilevel image coder/decoder system. While this system psrforms well on halftones, it similarly would not operate as a simple extension of the standard bilevel compression algorithms or other commonly-usQd techniqnas.
. Other examples sre found in U.S.Pat.No.4,571,634 to `~
CANESCHI ET AL uhich describas a coder/decoder for bilevel and halftoned ims~es-in which the hslftones axe formed by having areas of strict alternation of black and white p~ls, thus restricting its applicability;
U.S.Pat.No.4,425,582 tQ KADAKIA ET AL and U.S.Pat.No.4,435,726 to LIAO which disclose an efficient hardware implementation for a specific predictor that ls designsd to work well on both text and halftone data ;
wherein the predictor and a corresponding te-predictor would be added to an encoder/decoder systam, such as ~ ;~
tha~ defined by the GCITT G3/G4 standards, to produce a version of the origlnal data altered by a predic~or scheme; and-U.S.Pat.No.4,193,096 to STOFFEL which shows Y09~9-001 - 3 _ 2[3~837~
a system that scans and halftones an image and then compresses the resulting image based on knowlsdge of the halftoning process used, so that compression is depend-ent on tha halftoning process generating the imsge dat~
and thus this teaching i~ inapplicable in a system where all that is supplied to the coder is the bilevel image,-ant the coder has no control over the ha}ftoning proc~
ess.
i Gonsequently, it is desirable, and an object of the present invention, to provide a simple and versatile technique that facilitates the ef~icient compression of binary halftone data representing continuous tone images by known and widely-uset compression processes. -.
SUMMARY OF THE INVENTION
The system and method of the present invention is di-rected to taking a binary halftoned image with a given halftone pattern frequency and reformatting it to obtain image data which compresses better than the original image data using the CCITT standard Group 3 or Group 4 two-dimensional compression techniques or other cur-rently popular data compression processes. Accordingly, each H lines of the original image srs concatenated into ~, - . -~
20(~8370 the form of a single image line`for purposes of com-pression, where H is the halftone pattern frequency. The resulting reformatted Lmage tata may then be compressed using one of the CCITT algorith~s, or with any othar algorithm which makes use of correlations between fea-tures on a current line and those on the immediately preceding line, in an efficient manner.
fur~her feature of the present invention is the de-termining, for an ~mags with unknown pattern frequency, of a good estimate of the frequency for use in refor-matting the image.
,. :
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a small piece of a typical halftoned image with a pattern fxequency of four, in both original and ~ ;~
raformattad form.
FIG. 2 contras~s the coding of a line of the original ~`
i~aga fragme~t using tha reference data from the (a) original ~nd (b) reformatted images.
FIG. 3 shows a method for estimating the halftone pat-tern frequency of an image.
Y098~-~01 - 5 -2~ i371:) DETAIT.~D DESCRIPTION OF A PREFERRED EMBO~IMENT OF THE
INVENTION
,~
The reformatting of ~ halftoned i~age in accordance with the present invention is illustrated in Figure 1. The ~ -~
upper image shows a small piece of a typical halftoned image. Esch square is one pel tpicture element). The small lines below and to the right of the image data indicate the demarcation between the pel rows and col- -u~ns. Ihe halftone pattern is formed Ln a 4x4 pel block.
The reformatted image is shown below with lines 1, 2,
3, and 4 having been concatenatad to form the firs~ line;
lines 5, 6, 7, and 8 to form the second line; and suc-ces~ive groups of four lines forming the subsequent ~`
line-~. It will bc seen that the reformatted i~age shows ~;
much greater vertical correlation of edges from one line to the next. Facsimils COmpreSsiQn algorithms such as the CCITT Group 3 two-dimensional and Group 4 algorithms assu~e that such correlation is typical of bilevel im-ages, and have relatively shor~ code words to tescribe the existence of such correlation. Thus, if these com-pression algorithms are applied to an image which shows greatar correlation, better compression will result.
For example, the upper original image shown in Figure 1 requires 88 bytes to code using the CCITT Group 4 algo-20~337~) rithm, while ~he lower reformatted image requires only 46 bytes to store the same information. Reformatting a halftoned image as described allows each feature to be coded in a contsxt of ~eatures from the same position S Ln the halftone pattern above it. rhis is of significant value when many common for~s of haiftoning such 8S
supercircle or dither have been used to create the image to be compressed; it is of les~er value with halftoning tech~iques such as error diffusion which do not create any regular pa~tern.
Figure 2 illustrates the advantage to be 8ained in cod-Ing reformatted halftone data as compared to unreformatted data. Both portions, (a) and tb), of th¢
figurc show coding of line 9 of the original ;;~
unreformattted image of Figure l. Lina 9 forms the second line in each illustration in Figure 2, while the first line shows the reference data used for coding. ~-In Figure 2a, the reference line is line 8 of the ori-ginal ima~e, which would be used as the reference data in coding the unreformatted image. In the reformatted image, the reference data will be from the line in the original image locatPd four lines back from the line being coded, i.e. line 5, 8S shown in Figure Zb. It will ba understood that in Figure 2b alI of the reformatted YO989 0Ql - t -20(~3~0 image lines are not shown, only the parts which corre-spond to the tata in Figure 2a.
Below each image fragment i~ Figure 2 the CCITT G4 bit patterns which encote the image data are shown, along with a shorthand interpretation of the meanings of the codes. In G4 each black/white or white/black transi~ion causes one or (occasionally) more coding operations.
In Figure 2a, the first white run end coded lies two pels to the right of the reference run end (the left edge of ;~
the image in this case). This is indicated by a six-bit "vertical right two" code. The naxt two transitions define a white run which does not correspond to any run ~-~
on the upper reference or history line, so a "run-length" prefix is used a~d the two runs ~a two-pel black run and a two-pel white run) are coded using the one-dimensio~al codes. Following this, there is a black run ;~
which ends one pel to the left of the corresponding run on the raference line, so a three-bit "vertical left one" code ~s used. The next white run lines up exactly with the reference data, producing a one-bit "vertical zaro" code. This pattern ~VLl,V0) is repeated three more times for the next halftone pattern blocks. The final black run is aligned with the reference black run on ths right edge of the image, so another "vertical 2C3~83'~0 ~
zero" code is produced. Thus, the 24-bit image line has been "compressed" to 32 bits.
, In contrast, th~ same line can be coded in 16 bits by using the reference data provided by the reformatted S image, as shown in Figure 2b. The first black run on ~`
the reference line does not correspond to anything on ~ -the line beLng coded, so a "pass" code is genQrated.
After that all of the run ends are exactly aligned with the reference run ends, so a one-bit "vertical zero" i5 coded for each ~m end. In this case the reformatting has mad~ the difference between expansion snd modes~
compression of the halftoned data. In halftoned areas ;
which are very light or very dark, the compression may ` -~
be greater because not all lines have runs of both black and white in every halftone block, so that somewhat longer runs occur in the reformatted image.
. .
In general then, for a half~oned image consis~ing of R ;
rows and C columns and having a known pattern fr~quency N, compression can often be improved by reformatting tbe image to create an image having R/H rows and CxH columns, in ~hich the first H rows of the original image are concatenated in sequence to form the first row of the , Y0989-001 - 9 - ~
20~837~
reformatted image, the next H rows of the original image are concatenated to form the second row of the refor-matted image, and successive group~ of H rows accord-ingly form the subsequent rows.
5 . In practice, reformatting of an imàge may be done in various waysl but a simple technique to achiave it im-plicitly is ~y specifying an altered set of parameters to the encoder. For example, in an encoding system such -`
as thar described in U.S.Pat.No.4,725,815 issued Febru-ary 16, 1988 to MITCHELL, ANDERSON, and GOERTZEL, enti-tled "Method for Encoding and Decoding a Digital Image", ~herein tha coder is embodied in a computer program, one of the values which is input into the program is ~he number of pels per image line. If the lines of the image ~-~
: :
to b~ coded are arranged sequentially in storage for input into the program, the true number of pels per image line can be replaced with the product of the number of pels per ima~e lina and the halftone pattern frequency, thus in effect producing the desired image reforma~tin~.
More particularly, if there are 1000 pels per line and the pattern fraquency is 4, the number of pels per line could be presented to the encoder as 4000. Under thes~
circumstances, the decoder would have to decode the im-age assuming t~e sa~e number of pels per line which was Y0989~001 - 10 -~``` 2~837~
supplied to the encoder. It would then reformat the resulting image, which is a trivial operation when the decoded image lines are arranged sequentially in stor-age.
S In many inst~nces the halftone pattern frequency of a given image will be known, since the image will have been halftoned by a known system or process that produces halftone patterns of a particular frequency. However, this is not always the case. If the pattern Prequency is unknownl it becomes desirable to be able to quickly determine a workable value in a simple manner such as by selecting a~ estimated value based on an examination of the image. Conceptually, one way to do this is to compare each lina with each of the previous L history j lines, where L is some appropriate maximum pattern fre-quency value. For e~ample, it might be assu~ed that the pattern frequency will be no more tha~ 8, i.e., L = 8.
~fter doing the comparisons with the preceding 8 lines for one selected image line, ~he history line which produced the best match is chosen and its position (e.g.
"4th line back") is used as an estimate E of the halftone frequency for that line. By processing each line in the image in this way, a set of estimated frequencies E can Z0~8370 be developed. The one which occurs most often would be selected as the frequency to be used for~reformatting.
It ~ill be seen thst if an image processed in this way is actually a "true bilevel" image (e.g., text or line S art), the bast correlation uill normally be with the immediately preceding iine, and so a "pattern frequency"
of 1 will result and the image will be &oded in its or- ; -iginal format.
A detalled description of a preferred embodiment for carrying out an estimating procedure in accordance with the inventlon, particularly in the form of a program ~o ~ ;
run on a general-purpose computer, such as an I~M
System/370, will now be set forth with refarenca to Figure 3.
The flowchart of Figure 3 show$ a method for eseima~ing the halftone pattern frequency of a typical image to be reformatted in accordance ~ith the present invention.
The maximum frequency, MAXFRQ, to be used in the search for the estimated reformatting frequency, FREQ, is ini-tially selected os estimated based on programming, hardware, or execution speed considerations. For a given or current line~ the preceding lines sre examined as far back as the preset distance L, mentionet above (L =
~837~ :
MAXFRQ), to find the best ma~ch. Of course, the more lines that are examined, the more time it will take to do the comparisons, and the more storage that will have to be used to keep all of the required previous linQs S availabla. Consequently, it is desirable not to do a comparison with every line going back to thc beginnin8 of the image. It accordingly should ba determined how far back it is worthwhile to look for a match, based on ;~
the above-noted considerations for the system in which the invention is being applied, in particular, e.g., -~
uhat size hnlftone patterns are likely to appear and how much processing effort is to be expended to find the pattern value. Clearly, iP a large MAXFRQ is chosen and the pattern frequency is small, considerable time is `
wasted comparing lines that are too far back to be of interest. In attemp~ing to circumvent this problem one `
might design a system in ~hich successive lines begin-ning with the preceding line and moving backward through the image are examined in turn, with the comparison stoppin~ as soon as a line which is not a better match than the previously examined line is found. With this approach, however~ the sys~em may get cau~ht in a local minimum and "give up" before finding a much better match a little far~her back. Of course, one circ~mstance in which the search might be aborted before all lines have , Y~989-001 - 13 20(~837~
been compared is when an exact match for the current line is found. Otherwise, it is certainly desirable to exam-in~ all lines which are reasonable candidates for a match. MAXFRQ is thus selected to be the minimum dis-tance estimated to be needed to look back to pick up a reliable pattern frequency. For most images of the type contemplated, i.e., about 240 pel/inch, the pa~tern frequency is usually rather small, e.g., 4 or ~; but for a 600 pel/inch scanned halftone it may be about 11 or 12.
Once the ~XFRQ has been ~elected, the vector MATCH
keeps track of how many times each frequency fro~ 1 to MAXFRQ appears to be the best estimate of the pattern requency for an imaga line, based on two different `~
criteria: 1) the number of pels which differ between the line being examined and a reference or history line; and 2) the number of white/black and black/white trsnsitions on the line formed by exclusive-ORing the line being examined and a history line. Two criteria are used in-stesd of one because neither is a perfect estimator of the similarity between lines. By way cf explanation, generally, if two history lines produce about the ssme -~
number of runs on ~he exclusive-OR'd line, the line havi~g fewer pels different is th~ better match for the Y0989-001 - ~4 -2~83'7(~
line being examined; but, the number of pels which dif-fer does not always indicate bettes correlation. For example, two lines wh$ch are identical except for one - very long run added to one will have many pels which S differ, but only two run ends in the exclusive-OR'd line, and only a relatively small amount of additional data will bs required to code one with reference to th2 o~her. Similarly, a line on which sevaral edges shift by a single pel will have very few pels different from a rsference line, but many added transitions on the exclusive-OR'd line, and a large amount o~ additional data will be coded.
The large loop on the left side of the chart estimates the pattern frequency E for each line in sequence. MATCH ~ `~
is zeroed initially. MAXFRQ blank lines are stored in ;
a history buffer, so that there will be enough history lines to compare to the initial image lines. At the begir~ing of the loop, if another input line remains, it is read in. An index I is initialized to 1, and the minim~m different pel count DMIN and the minimum tran-sition count XMIN are initialized to the number of pels per line NCOLS plus one, so that comparison with at least one line must find fewer pels different and fewer tran-sitions. A loop is then entered which compares the cur-20~8370 rent line to the I'th line back, where I runs from 1 to MAXFRQ. For each of the I history lines, the ~ `
exclusi~e-OR of the current line and the history lina is formet. Tha number D of l-valued pels in the exclusive-OR'd line gives the numbsr of pels which dif-fer between tha two lines; and if this value is less than DMIN, ie replaces DMIN, and the current value of I is saved as DI. The number X of transitions in the exclusive-OR'd line is compared to XMIN; and if the new value is smaller, it replaces XMIN, and the value of I
is saved as XI. It should be noted that if D=DMIN or -;
X=XMIN, the minimum value is not updated; this resolves "ties" in favor of the line closer to the current line.
When the smaller loop is exited, Dl and XI contain es-timates E of the pattern frequency for the current line based on the two difference criteria. The counters ~ ~`
MATGH(DI) and MATCH(XI) are incremented to record "votes" for those frequencies, and the loop repeats to process the next line. `~
After all of the image lines have been processed in this m~nner, MATGH is examined to determine which frequency, among the Es, was most often chosen as ehe likely pattern frequency. The maximum number of times a value has besn chosen, MAXCT, is initializet to zero, and ~ loop is YOg8g-001 - 16 - -Z~ 3~0 entered which finds the maximum value of MATCH(I) where I runs from 1 to ~AXFRQ. This value is saved as MAXCT, ~ -and the value of I which corresponds to it is saved as FREQ. On exit from thi~ loop, FREQ gives the estimate `~
of *he halftone pattern frequency, ~hat is, the number of lines ~ which will be concatenated in accordance with ~;
the present invention.
, It will be appreciated that variatior.s on this general scheme are possible with~n the scope of the invention.
One of the criteria for deciding w~lich history line matches the current line best migh~ be omitted, or some alternative criteria may be added or substituted. Dif-fer~nt criteria could be given different weights in the decision-making process. For exampls, if the value of D
in the first loop is considered to give a more accurate indication of the quality of th~ match than X, then MATCH(DI) could be incremented by a greater value than -MATCH(XI); the DI values would then dominate the deci-sion made in the final loop. The invention might be extended to graylevel or color ~mages in situations where a pattern frequency may exist (e.g. for a scanned halftone), snt it might also be applied to portions of an imageJ allowing some switching betwePn pattern fre-quencies when part o an image consists of halftone data Y0989-~01 - 17 -.
Z~83~ ~
and part consists of more conventional facsimile mate-rial (text and line art).
It will accordingly be seen that a system and method have been disclosed for obtaining good compression of halftone images crea~ed with a known halftone pattern frequency. An original bilevel image is reformatted to produce another bilevel image that allows vertical cor-relations to be recognized by the compression technique, thus improving compressibility dramatically, with par-ticular suitability for facsimile transmissions. The reformatting is carried out by concatenating each suc cessive group of H lines together (where H ls the hslftone pattern frequency), ant using the CCITT Group 3 two-dimensional (MR) algorithm, the CCITT Group 4 al-gorithm (MMR), I~M ~DNR, or any similar two-dimensional coding procedure to encode the reformatted image. Fur-ther, the halftone pattern frequency of an image of un-known characteristics can readily be estimated by examination of the ima8e in the manner described.
lines 5, 6, 7, and 8 to form the second line; and suc-ces~ive groups of four lines forming the subsequent ~`
line-~. It will bc seen that the reformatted i~age shows ~;
much greater vertical correlation of edges from one line to the next. Facsimils COmpreSsiQn algorithms such as the CCITT Group 3 two-dimensional and Group 4 algorithms assu~e that such correlation is typical of bilevel im-ages, and have relatively shor~ code words to tescribe the existence of such correlation. Thus, if these com-pression algorithms are applied to an image which shows greatar correlation, better compression will result.
For example, the upper original image shown in Figure 1 requires 88 bytes to code using the CCITT Group 4 algo-20~337~) rithm, while ~he lower reformatted image requires only 46 bytes to store the same information. Reformatting a halftoned image as described allows each feature to be coded in a contsxt of ~eatures from the same position S Ln the halftone pattern above it. rhis is of significant value when many common for~s of haiftoning such 8S
supercircle or dither have been used to create the image to be compressed; it is of les~er value with halftoning tech~iques such as error diffusion which do not create any regular pa~tern.
Figure 2 illustrates the advantage to be 8ained in cod-Ing reformatted halftone data as compared to unreformatted data. Both portions, (a) and tb), of th¢
figurc show coding of line 9 of the original ;;~
unreformattted image of Figure l. Lina 9 forms the second line in each illustration in Figure 2, while the first line shows the reference data used for coding. ~-In Figure 2a, the reference line is line 8 of the ori-ginal ima~e, which would be used as the reference data in coding the unreformatted image. In the reformatted image, the reference data will be from the line in the original image locatPd four lines back from the line being coded, i.e. line 5, 8S shown in Figure Zb. It will ba understood that in Figure 2b alI of the reformatted YO989 0Ql - t -20(~3~0 image lines are not shown, only the parts which corre-spond to the tata in Figure 2a.
Below each image fragment i~ Figure 2 the CCITT G4 bit patterns which encote the image data are shown, along with a shorthand interpretation of the meanings of the codes. In G4 each black/white or white/black transi~ion causes one or (occasionally) more coding operations.
In Figure 2a, the first white run end coded lies two pels to the right of the reference run end (the left edge of ;~
the image in this case). This is indicated by a six-bit "vertical right two" code. The naxt two transitions define a white run which does not correspond to any run ~-~
on the upper reference or history line, so a "run-length" prefix is used a~d the two runs ~a two-pel black run and a two-pel white run) are coded using the one-dimensio~al codes. Following this, there is a black run ;~
which ends one pel to the left of the corresponding run on the raference line, so a three-bit "vertical left one" code ~s used. The next white run lines up exactly with the reference data, producing a one-bit "vertical zaro" code. This pattern ~VLl,V0) is repeated three more times for the next halftone pattern blocks. The final black run is aligned with the reference black run on ths right edge of the image, so another "vertical 2C3~83'~0 ~
zero" code is produced. Thus, the 24-bit image line has been "compressed" to 32 bits.
, In contrast, th~ same line can be coded in 16 bits by using the reference data provided by the reformatted S image, as shown in Figure 2b. The first black run on ~`
the reference line does not correspond to anything on ~ -the line beLng coded, so a "pass" code is genQrated.
After that all of the run ends are exactly aligned with the reference run ends, so a one-bit "vertical zero" i5 coded for each ~m end. In this case the reformatting has mad~ the difference between expansion snd modes~
compression of the halftoned data. In halftoned areas ;
which are very light or very dark, the compression may ` -~
be greater because not all lines have runs of both black and white in every halftone block, so that somewhat longer runs occur in the reformatted image.
. .
In general then, for a half~oned image consis~ing of R ;
rows and C columns and having a known pattern fr~quency N, compression can often be improved by reformatting tbe image to create an image having R/H rows and CxH columns, in ~hich the first H rows of the original image are concatenated in sequence to form the first row of the , Y0989-001 - 9 - ~
20~837~
reformatted image, the next H rows of the original image are concatenated to form the second row of the refor-matted image, and successive group~ of H rows accord-ingly form the subsequent rows.
5 . In practice, reformatting of an imàge may be done in various waysl but a simple technique to achiave it im-plicitly is ~y specifying an altered set of parameters to the encoder. For example, in an encoding system such -`
as thar described in U.S.Pat.No.4,725,815 issued Febru-ary 16, 1988 to MITCHELL, ANDERSON, and GOERTZEL, enti-tled "Method for Encoding and Decoding a Digital Image", ~herein tha coder is embodied in a computer program, one of the values which is input into the program is ~he number of pels per image line. If the lines of the image ~-~
: :
to b~ coded are arranged sequentially in storage for input into the program, the true number of pels per image line can be replaced with the product of the number of pels per ima~e lina and the halftone pattern frequency, thus in effect producing the desired image reforma~tin~.
More particularly, if there are 1000 pels per line and the pattern fraquency is 4, the number of pels per line could be presented to the encoder as 4000. Under thes~
circumstances, the decoder would have to decode the im-age assuming t~e sa~e number of pels per line which was Y0989~001 - 10 -~``` 2~837~
supplied to the encoder. It would then reformat the resulting image, which is a trivial operation when the decoded image lines are arranged sequentially in stor-age.
S In many inst~nces the halftone pattern frequency of a given image will be known, since the image will have been halftoned by a known system or process that produces halftone patterns of a particular frequency. However, this is not always the case. If the pattern Prequency is unknownl it becomes desirable to be able to quickly determine a workable value in a simple manner such as by selecting a~ estimated value based on an examination of the image. Conceptually, one way to do this is to compare each lina with each of the previous L history j lines, where L is some appropriate maximum pattern fre-quency value. For e~ample, it might be assu~ed that the pattern frequency will be no more tha~ 8, i.e., L = 8.
~fter doing the comparisons with the preceding 8 lines for one selected image line, ~he history line which produced the best match is chosen and its position (e.g.
"4th line back") is used as an estimate E of the halftone frequency for that line. By processing each line in the image in this way, a set of estimated frequencies E can Z0~8370 be developed. The one which occurs most often would be selected as the frequency to be used for~reformatting.
It ~ill be seen thst if an image processed in this way is actually a "true bilevel" image (e.g., text or line S art), the bast correlation uill normally be with the immediately preceding iine, and so a "pattern frequency"
of 1 will result and the image will be &oded in its or- ; -iginal format.
A detalled description of a preferred embodiment for carrying out an estimating procedure in accordance with the inventlon, particularly in the form of a program ~o ~ ;
run on a general-purpose computer, such as an I~M
System/370, will now be set forth with refarenca to Figure 3.
The flowchart of Figure 3 show$ a method for eseima~ing the halftone pattern frequency of a typical image to be reformatted in accordance ~ith the present invention.
The maximum frequency, MAXFRQ, to be used in the search for the estimated reformatting frequency, FREQ, is ini-tially selected os estimated based on programming, hardware, or execution speed considerations. For a given or current line~ the preceding lines sre examined as far back as the preset distance L, mentionet above (L =
~837~ :
MAXFRQ), to find the best ma~ch. Of course, the more lines that are examined, the more time it will take to do the comparisons, and the more storage that will have to be used to keep all of the required previous linQs S availabla. Consequently, it is desirable not to do a comparison with every line going back to thc beginnin8 of the image. It accordingly should ba determined how far back it is worthwhile to look for a match, based on ;~
the above-noted considerations for the system in which the invention is being applied, in particular, e.g., -~
uhat size hnlftone patterns are likely to appear and how much processing effort is to be expended to find the pattern value. Clearly, iP a large MAXFRQ is chosen and the pattern frequency is small, considerable time is `
wasted comparing lines that are too far back to be of interest. In attemp~ing to circumvent this problem one `
might design a system in ~hich successive lines begin-ning with the preceding line and moving backward through the image are examined in turn, with the comparison stoppin~ as soon as a line which is not a better match than the previously examined line is found. With this approach, however~ the sys~em may get cau~ht in a local minimum and "give up" before finding a much better match a little far~her back. Of course, one circ~mstance in which the search might be aborted before all lines have , Y~989-001 - 13 20(~837~
been compared is when an exact match for the current line is found. Otherwise, it is certainly desirable to exam-in~ all lines which are reasonable candidates for a match. MAXFRQ is thus selected to be the minimum dis-tance estimated to be needed to look back to pick up a reliable pattern frequency. For most images of the type contemplated, i.e., about 240 pel/inch, the pa~tern frequency is usually rather small, e.g., 4 or ~; but for a 600 pel/inch scanned halftone it may be about 11 or 12.
Once the ~XFRQ has been ~elected, the vector MATCH
keeps track of how many times each frequency fro~ 1 to MAXFRQ appears to be the best estimate of the pattern requency for an imaga line, based on two different `~
criteria: 1) the number of pels which differ between the line being examined and a reference or history line; and 2) the number of white/black and black/white trsnsitions on the line formed by exclusive-ORing the line being examined and a history line. Two criteria are used in-stesd of one because neither is a perfect estimator of the similarity between lines. By way cf explanation, generally, if two history lines produce about the ssme -~
number of runs on ~he exclusive-OR'd line, the line havi~g fewer pels different is th~ better match for the Y0989-001 - ~4 -2~83'7(~
line being examined; but, the number of pels which dif-fer does not always indicate bettes correlation. For example, two lines wh$ch are identical except for one - very long run added to one will have many pels which S differ, but only two run ends in the exclusive-OR'd line, and only a relatively small amount of additional data will bs required to code one with reference to th2 o~her. Similarly, a line on which sevaral edges shift by a single pel will have very few pels different from a rsference line, but many added transitions on the exclusive-OR'd line, and a large amount o~ additional data will be coded.
The large loop on the left side of the chart estimates the pattern frequency E for each line in sequence. MATCH ~ `~
is zeroed initially. MAXFRQ blank lines are stored in ;
a history buffer, so that there will be enough history lines to compare to the initial image lines. At the begir~ing of the loop, if another input line remains, it is read in. An index I is initialized to 1, and the minim~m different pel count DMIN and the minimum tran-sition count XMIN are initialized to the number of pels per line NCOLS plus one, so that comparison with at least one line must find fewer pels different and fewer tran-sitions. A loop is then entered which compares the cur-20~8370 rent line to the I'th line back, where I runs from 1 to MAXFRQ. For each of the I history lines, the ~ `
exclusi~e-OR of the current line and the history lina is formet. Tha number D of l-valued pels in the exclusive-OR'd line gives the numbsr of pels which dif-fer between tha two lines; and if this value is less than DMIN, ie replaces DMIN, and the current value of I is saved as DI. The number X of transitions in the exclusive-OR'd line is compared to XMIN; and if the new value is smaller, it replaces XMIN, and the value of I
is saved as XI. It should be noted that if D=DMIN or -;
X=XMIN, the minimum value is not updated; this resolves "ties" in favor of the line closer to the current line.
When the smaller loop is exited, Dl and XI contain es-timates E of the pattern frequency for the current line based on the two difference criteria. The counters ~ ~`
MATGH(DI) and MATCH(XI) are incremented to record "votes" for those frequencies, and the loop repeats to process the next line. `~
After all of the image lines have been processed in this m~nner, MATGH is examined to determine which frequency, among the Es, was most often chosen as ehe likely pattern frequency. The maximum number of times a value has besn chosen, MAXCT, is initializet to zero, and ~ loop is YOg8g-001 - 16 - -Z~ 3~0 entered which finds the maximum value of MATCH(I) where I runs from 1 to ~AXFRQ. This value is saved as MAXCT, ~ -and the value of I which corresponds to it is saved as FREQ. On exit from thi~ loop, FREQ gives the estimate `~
of *he halftone pattern frequency, ~hat is, the number of lines ~ which will be concatenated in accordance with ~;
the present invention.
, It will be appreciated that variatior.s on this general scheme are possible with~n the scope of the invention.
One of the criteria for deciding w~lich history line matches the current line best migh~ be omitted, or some alternative criteria may be added or substituted. Dif-fer~nt criteria could be given different weights in the decision-making process. For exampls, if the value of D
in the first loop is considered to give a more accurate indication of the quality of th~ match than X, then MATCH(DI) could be incremented by a greater value than -MATCH(XI); the DI values would then dominate the deci-sion made in the final loop. The invention might be extended to graylevel or color ~mages in situations where a pattern frequency may exist (e.g. for a scanned halftone), snt it might also be applied to portions of an imageJ allowing some switching betwePn pattern fre-quencies when part o an image consists of halftone data Y0989-~01 - 17 -.
Z~83~ ~
and part consists of more conventional facsimile mate-rial (text and line art).
It will accordingly be seen that a system and method have been disclosed for obtaining good compression of halftone images crea~ed with a known halftone pattern frequency. An original bilevel image is reformatted to produce another bilevel image that allows vertical cor-relations to be recognized by the compression technique, thus improving compressibility dramatically, with par-ticular suitability for facsimile transmissions. The reformatting is carried out by concatenating each suc cessive group of H lines together (where H ls the hslftone pattern frequency), ant using the CCITT Group 3 two-dimensional (MR) algorithm, the CCITT Group 4 al-gorithm (MMR), I~M ~DNR, or any similar two-dimensional coding procedure to encode the reformatted image. Fur-ther, the halftone pattern frequency of an image of un-known characteristics can readily be estimated by examination of the ima8e in the manner described.
Claims (10)
1. A method of compressing an image having a pattern frequency H comprising:
reformatting said image by concatenating the first H lines in sequence to form a single line, con cat-enating the next H lines to form 8 second line, and repeating this process of concatenating lines until the entire image has been so reformatted; and applying a compression algorithm which makes use of directional correlations to compress said re-formatted image.
reformatting said image by concatenating the first H lines in sequence to form a single line, con cat-enating the next H lines to form 8 second line, and repeating this process of concatenating lines until the entire image has been so reformatted; and applying a compression algorithm which makes use of directional correlations to compress said re-formatted image.
2. A system for compressing an image having a pattern frequency H comprising:
means for reformatting said image by concatenating the first H lines in sequence to form a single line, concatenating the next H lines to form a second line, and repeating the concatenating in this man-ner until the entire image has been so reformatted;
and means for applying a compression algorithm which makes use of directional correlations to compress said reformatted image.
means for reformatting said image by concatenating the first H lines in sequence to form a single line, concatenating the next H lines to form a second line, and repeating the concatenating in this man-ner until the entire image has been so reformatted;
and means for applying a compression algorithm which makes use of directional correlations to compress said reformatted image.
3. A system for determining the pattern frequency H of an image comprising:
means for selecting a maximum pattern frequency MAXFREQ to be used in searching for said pattern frequency H;
means for comparing a given line of said image with each of MAXFREQ preceding lines of said image, means for determining which preceding line of said MAXFREQ preceding lines most closely matches said given line;
means for selecting the number E of lines before said given line at which the best match occurs to be the estimated halftone pattern frequency;
means for repeatedly actuating said comparing, de-termining, and selecting means with successive im-age lines being the given line to compute a plurality of E values for said image; and means for selecting the most frequently occurring E value as the pattern frequency H of the image.
means for selecting a maximum pattern frequency MAXFREQ to be used in searching for said pattern frequency H;
means for comparing a given line of said image with each of MAXFREQ preceding lines of said image, means for determining which preceding line of said MAXFREQ preceding lines most closely matches said given line;
means for selecting the number E of lines before said given line at which the best match occurs to be the estimated halftone pattern frequency;
means for repeatedly actuating said comparing, de-termining, and selecting means with successive im-age lines being the given line to compute a plurality of E values for said image; and means for selecting the most frequently occurring E value as the pattern frequency H of the image.
4. A system as in claim 3 wherein said image is a bilevel image and said comparing means comprises:
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining means comprises:
means for counting the number of l-value and O-value runs in each resultant line; and means for noting the number of lines E before said given line at which the count of runs first reaches a minimum value.
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining means comprises:
means for counting the number of l-value and O-value runs in each resultant line; and means for noting the number of lines E before said given line at which the count of runs first reaches a minimum value.
5. A system as in claim 3 wherein said image is a bilevel image and said comparing means comprises:
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining means comprises:
means for counting the number of 1-valued pels in each resultant line; and means for noting the number of lines E before said given line at which the count of 1-valued pels first reaches a minimum value.
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining means comprises:
means for counting the number of 1-valued pels in each resultant line; and means for noting the number of lines E before said given line at which the count of 1-valued pels first reaches a minimum value.
6. A method for determining the pattern frequency H of an image comprising:
selecting a maximum pattern frequency MAXFREQ to be used in searching for said pattern frequency H;
comparing a given line of said image with each of MAXFREQ preceding lines of said image;
determining which preceding line of said MAXFREQ
preceding lines most closely matches said given line;
selecting the number E of lines before said given line at which the best match occurs to be the es-timated pattern frequency;
repeating said comparing, determining, and select-ing steps with successive image lines being the given line to compute a plurality of E values for said image; and selecting the most frequently occurring E value as the pattern frequency H of the image.
selecting a maximum pattern frequency MAXFREQ to be used in searching for said pattern frequency H;
comparing a given line of said image with each of MAXFREQ preceding lines of said image;
determining which preceding line of said MAXFREQ
preceding lines most closely matches said given line;
selecting the number E of lines before said given line at which the best match occurs to be the es-timated pattern frequency;
repeating said comparing, determining, and select-ing steps with successive image lines being the given line to compute a plurality of E values for said image; and selecting the most frequently occurring E value as the pattern frequency H of the image.
7. A method as in claim 6 wherein said image is a bilevel image and said comparing step comprises:
exclusive-ORing the pal values In a given line of said original image with the respective pel values.
in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining step comprises:
counting the number of l-value and 0-value runs in each resultant line; and noting the number of lines E before said given line at which the count of runs first reaches a minimum value.
exclusive-ORing the pal values In a given line of said original image with the respective pel values.
in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining step comprises:
counting the number of l-value and 0-value runs in each resultant line; and noting the number of lines E before said given line at which the count of runs first reaches a minimum value.
8. A method as in claim 6 wherein said image is a bilevel image and said comparing step comprises:
exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining step comprises:
counting the number of 1-valued pels in each re-sultant line; and noting the number of lines E before said given line at which the count of 1-valued pels first reaches a minimum value.
exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of MAXFREQ preceding lines of said image to produce respective resultant lines;
and said determining step comprises:
counting the number of 1-valued pels in each re-sultant line; and noting the number of lines E before said given line at which the count of 1-valued pels first reaches a minimum value.
19. A system for reformatting halftone data wherein an original bilevel image is reformatted to get an-other bilevel image which allows vertical corre-lations to be recognized comprising:
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of a selected number of preceding lines of said image to produce respective resultant lines;
means for counting the number of 1-value and 0-value runs in each resultant line;
means for counting the number of 1-valued pels in each resultant line;
means for determining the respective number of lines E preceding said given line at which the count of runs and the count of l-valued pels first reach minimum values;
means for storing said E values as estimated halftone pattern frequencies;
means for repeatedly actuating each of the preced-ing means with successive lines being the given line to compute a plurality of E values;
means for selecting the most frequently occurring E value as the halftone pattern frequency H of the image; and means for reformatting said image by concatenating the first H lines in sequence to form a single line, concatenating the next H lines in sequence to form a second line, and repeating the concatenating in this manner until the entire image has been so re-formatted.
means for exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of a selected number of preceding lines of said image to produce respective resultant lines;
means for counting the number of 1-value and 0-value runs in each resultant line;
means for counting the number of 1-valued pels in each resultant line;
means for determining the respective number of lines E preceding said given line at which the count of runs and the count of l-valued pels first reach minimum values;
means for storing said E values as estimated halftone pattern frequencies;
means for repeatedly actuating each of the preced-ing means with successive lines being the given line to compute a plurality of E values;
means for selecting the most frequently occurring E value as the halftone pattern frequency H of the image; and means for reformatting said image by concatenating the first H lines in sequence to form a single line, concatenating the next H lines in sequence to form a second line, and repeating the concatenating in this manner until the entire image has been so re-formatted.
10. A method for reformatting halftone data wherein an original bilevel image is reformatted to get an-other bilevel image which allows vertical corre-lations to be recognized comprising the steps of:
exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of a selected number of preceding lines of said image to produce respective resultant lines;
counting the number of 1-value and 0-value runs in each resultant line;
counting the number of 1-valued pels in each re-sultant line;
determining the respective number of lines E pre-ceding said given line at which the count of runs and the count of l-valued pels first reach minimum values;
storing said E values as estimated halftone pattern frequencies;
repeating each of the preceding steps with succes-sive lines being the given line to compute a plu-rality of E values;
selecting the most frequently occurring E value as the halftone pattern frequency H of the image; and reformatting said image by concatenating the first H lines in sequence to form a single line, concat-enating the next H lines in sequence to form a second line, and repeating the concatenating in this manner until the entire image has been so re-formatted.
exclusive-ORing the pel values in a given line of said original image with the respective pel values in each of a selected number of preceding lines of said image to produce respective resultant lines;
counting the number of 1-value and 0-value runs in each resultant line;
counting the number of 1-valued pels in each re-sultant line;
determining the respective number of lines E pre-ceding said given line at which the count of runs and the count of l-valued pels first reach minimum values;
storing said E values as estimated halftone pattern frequencies;
repeating each of the preceding steps with succes-sive lines being the given line to compute a plu-rality of E values;
selecting the most frequently occurring E value as the halftone pattern frequency H of the image; and reformatting said image by concatenating the first H lines in sequence to form a single line, concat-enating the next H lines in sequence to form a second line, and repeating the concatenating in this manner until the entire image has been so re-formatted.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US07/305,313 US4965677A (en) | 1989-02-01 | 1989-02-01 | Compression of binary halftones |
| US07/305,313 | 1989-02-01 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CA2008370A1 CA2008370A1 (en) | 1990-08-01 |
| CA2008370C true CA2008370C (en) | 1995-02-07 |
Family
ID=23180291
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA002008370A Expired - Fee Related CA2008370C (en) | 1989-02-01 | 1990-01-23 | Compression of binary halftones |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US4965677A (en) |
| EP (1) | EP0381129A1 (en) |
| JP (1) | JPH02290371A (en) |
| CA (1) | CA2008370C (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE69032908T2 (en) * | 1989-11-20 | 1999-09-09 | Canon Kk | Image processing device |
| US5359430A (en) * | 1992-05-15 | 1994-10-25 | Microsoft Corporation | Block-halftoning method and system with compressed error image |
| US5864711A (en) * | 1995-07-05 | 1999-01-26 | Microsoft Corporation | System for determining more accurate translation between first and second translator, and providing translated data to second computer if first translator is more accurate |
| US6911987B1 (en) * | 1995-07-05 | 2005-06-28 | Microsoft Corporation | Method and system for transmitting data for a shared application |
| US6272256B1 (en) | 1997-11-26 | 2001-08-07 | International Business Machines Corporation | Fast compression of periodic halftoned bitonal images |
| US6798543B1 (en) | 2000-09-13 | 2004-09-28 | Sharp Laboratories Of America Inc. | Method and apparatus for compression and decompression of halftoned images |
| US6947604B2 (en) * | 2002-01-17 | 2005-09-20 | Intel Corporation | Method and hardware to implement two-dimensional compression |
| US7418664B2 (en) * | 2002-04-03 | 2008-08-26 | Microsoft Corporation | Application sharing single document sharing |
| US7028266B2 (en) | 2002-04-05 | 2006-04-11 | Microsoft Corporation | Processing occluded windows during application sharing |
| US8756513B1 (en) | 2002-04-23 | 2014-06-17 | Microsoft Corporation | Document viewing mechanism for document sharing environment |
| US7293243B1 (en) * | 2002-05-22 | 2007-11-06 | Microsoft Corporation | Application sharing viewer presentation |
| US7356563B1 (en) | 2002-06-06 | 2008-04-08 | Microsoft Corporation | Methods of annotating a collaborative application display |
| US7025515B2 (en) * | 2003-05-20 | 2006-04-11 | Software 2000 Ltd. | Bit mask generation system |
| US7005957B2 (en) * | 2004-05-29 | 2006-02-28 | Tsung-Mou Yu | Mechanism for trip-free of the bimetallic plate of a safety switch device |
| US7430064B2 (en) * | 2005-06-27 | 2008-09-30 | Xerox Corporation | Rotation dependent halftone rendering in scanning |
| US8848250B2 (en) * | 2012-10-23 | 2014-09-30 | Kyocera Document Solutions Inc. | Compression of halftoned digital images |
Family Cites Families (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE2233796C3 (en) * | 1972-07-10 | 1975-03-20 | K.K. Ricoh, Tokio | Method for video signal compression and expansion and devices for carrying out the method |
| GB1481226A (en) * | 1973-08-31 | 1977-07-27 | Kokusai Denshin Denwa Co Ltd | System for coding two-dimensional information |
| US4091424A (en) * | 1977-02-18 | 1978-05-23 | Compression Labs, Inc. | Facsimile compression system |
| US4213154A (en) * | 1978-08-03 | 1980-07-15 | Mitsubishi Denki Kabushiki Kaisha | Facsimile communication system |
| US4360840A (en) * | 1980-05-13 | 1982-11-23 | Am International, Inc. | Real time data compression/decompression scheme for facsimile transmission system |
| US4400738A (en) * | 1981-10-30 | 1983-08-23 | Xerox Corporation | Image signal processing method and apparatus |
| JPS58138176A (en) * | 1982-02-12 | 1983-08-16 | Kokusai Denshin Denwa Co Ltd <Kdd> | Encoding system for image having multilevel gradation |
| US4562467A (en) * | 1983-04-01 | 1985-12-31 | Itek Corporation | Data compression apparatus and method for encoding and decoding multi-line signals |
| GB2139849B (en) * | 1983-05-07 | 1986-11-19 | Nippon Telegraph & Telephone | Image data compression system |
| US4630125A (en) * | 1983-06-01 | 1986-12-16 | Xerox Corporation | Unscreening of stored digital halftone images |
| JPS60213171A (en) * | 1984-04-09 | 1985-10-25 | Ricoh Co Ltd | Encoding processor |
| US4729034A (en) * | 1985-06-27 | 1988-03-01 | Netexpress Systems, Inc. | Method and apparatus for selection of a coding reference line for two-dimensional coding of image data representing screened images |
| EP0241868B1 (en) * | 1986-04-16 | 1996-03-20 | Hitachi, Ltd. | Disk apparatus |
-
1989
- 1989-02-01 US US07/305,313 patent/US4965677A/en not_active Expired - Fee Related
-
1990
- 1990-01-23 CA CA002008370A patent/CA2008370C/en not_active Expired - Fee Related
- 1990-01-30 EP EP90101777A patent/EP0381129A1/en not_active Withdrawn
- 1990-02-01 JP JP2020555A patent/JPH02290371A/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| JPH02290371A (en) | 1990-11-30 |
| EP0381129A1 (en) | 1990-08-08 |
| US4965677A (en) | 1990-10-23 |
| CA2008370A1 (en) | 1990-08-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2008370C (en) | Compression of binary halftones | |
| EP0504903B1 (en) | Apparatus for and method of preprocessing binary picture data prior to run-length encoding | |
| US5297220A (en) | Image processing system for image compression and decompression | |
| EP0562420B1 (en) | Method and apparatus for vector coding of video transform coefficients | |
| US4888645A (en) | Method for encoding and decoding a digital image | |
| KR100530681B1 (en) | How to send and receive coded video images | |
| CA1064618A (en) | Compression of gray scale imagery to less than one bit per picture element | |
| US5583657A (en) | Method and apparatus for scanning image data | |
| JP4522199B2 (en) | Image encoding apparatus, image processing apparatus, control method therefor, computer program, and computer-readable storage medium | |
| US5113457A (en) | Image information code processing system | |
| JPH04328960A (en) | Picture data transmitter | |
| US5138673A (en) | Data compression system | |
| US6728412B1 (en) | Method and apparatus for on-the-fly image coding | |
| JPH03203463A (en) | Image processor | |
| EP0902398B1 (en) | Method and system for compressing and decompressing binary representations of dithered images | |
| JPH08125868A (en) | Method and device for processing picture | |
| CA1335309C (en) | Method for encoding and decoding a digital image | |
| US5406282A (en) | Data coding and decoding with improved efficiency | |
| JPH07264417A (en) | Image coding method | |
| CA2436437A1 (en) | Moving picture information compressing method and its system | |
| US4782400A (en) | System for encoding or decoding analog video signals | |
| US6084913A (en) | Method for compressing dynamic image information and system and device therefor | |
| JPH05151349A (en) | Image data compressing method and encoding circuit | |
| JPS63182973A (en) | Pseudo half-tonal image transmission method for facsimile equipment | |
| JP3295507B2 (en) | Binary image encoding / decoding method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |