WO2006022741A1 - Compression d'images utilisable avec des images animees - Google Patents

Compression d'images utilisable avec des images animees Download PDF

Info

Publication number
WO2006022741A1
WO2006022741A1 PCT/US2004/027389 US2004027389W WO2006022741A1 WO 2006022741 A1 WO2006022741 A1 WO 2006022741A1 US 2004027389 W US2004027389 W US 2004027389W WO 2006022741 A1 WO2006022741 A1 WO 2006022741A1
Authority
WO
WIPO (PCT)
Prior art keywords
color
image
pixel
compressed
probability
Prior art date
Application number
PCT/US2004/027389
Other languages
English (en)
Inventor
Robert M. Caruso
Konstantin J. Balashov
Original Assignee
Marger Johnson & Mccollom, P.C.
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Marger Johnson & Mccollom, P.C. filed Critical Marger Johnson & Mccollom, P.C.
Priority to PCT/US2004/027389 priority Critical patent/WO2006022741A1/fr
Publication of WO2006022741A1 publication Critical patent/WO2006022741A1/fr

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
    • H04N1/46—Colour picture communication systems
    • H04N1/64—Systems for the transmission or the storage of the colour picture signal; Details therefor, e.g. coding or decoding means therefor
    • H04N1/644—Systems for the transmission or the storage of the colour picture signal; Details therefor, e.g. coding or decoding means therefor using a reduced set of representative colours, e.g. each representing a particular range in a colour space
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103—Selection of coding mode or of prediction mode
    • H04N19/105—Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/182—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a pixel
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques

Definitions

  • This invention pertains to compression techniques on a computer, and more particularly to a lossless compression technique usable with digital color images, including animated images.
  • File compression technology has recently experienced a resurgence. Originally, file compression was necessary because disk space was limited. To maximize the number of files that could be stored on a disk, it was occasionally necessary to compress files. More recently, hard disk space has become very cheap, and users have been able to store massive quantities of data. The need for compression to save disk space has diminished.
  • One technique used to reduce the size of the file is to limit the number of colors used in the image. There are two reasons why including a large number of colors in an image is impractical or unnecessary. First, the computer hardware on which the image is displayed (i.e., the monitor and video card) might be limited in the number of colors that can be displayed at one time. Second, the human eye is limited in the number of colors it can distinguish when looking at an image. To address these concerns, a typical image uses a color palette, which includes either a subset of the colors in the image or approximations of the colors in the image. The number of entries in the color palette determines the number of different colors that occur in the image.
  • the color palette of the image includes 256 colors, but a person skilled in the art will recognize that this number can vary.
  • the Median Cut or a similar algorithm can be used to select the colors stored in the color palette. The specifics of how the colors are selected for the color palette is not relevant to the invention and will not be discussed here.
  • Using a color palette begins the process of compressing the image. For example, if the image is stored using 24-bit color, it takes three bytes to store the color for each pixel. If only 256 colors are used and stored in the color palette, the color for each pixel can be represented using only one byte: the index into the color palette. This reduces the size of the image file by two thirds.
  • a Huffman coding can be applied to the indices into the color palette for the pixels.
  • the frequencies for each symbol (in this case, the different colors in the color palette) in the message (in this case, the image) are calculated.
  • the entire image is scanned, and the number of times each color is counted is scanned.
  • the frequency for each color can then be determined by dividing the number of occurrences of each color by the total number of pixels in the image.
  • FIG. 10 shows the construction of the Huffman tree.
  • the frequencies start out as leaves 1005, 1010, 1015, and 1020 in a to-be- constructed tree.
  • the two smallest frequencies are assigned a common parent node in the tree, and the parent node is assigned a frequency equal to the sum of its children.
  • the two smallest frequencies are 0.10 and 0.20, which combine to a parent node frequency of 0.30.
  • the process then repeats, using the parent node's frequency in place of its two children, until only a single (root) node remains.
  • the two children of each parent node are assigned a "0" or a "1", depending on whether they are a "left” or a "right” branch from the parent node. (A person skilled in the art will recognize that the determination of which branch is "left” and which is "right” is arbitrary.)
  • the Huffman coding for each symbol is the sequence of branches from the root node of the Huffman tree to the leaf for that symbol. For example, the Huffman coding for symbol "D” is “1", whereas the Huffman coding for symbol "B” is "001.”
  • JPEG Joint Photographic Expert Group
  • MPEG Motion Picture Expert Group
  • JPEG and MPEG compression techniques are lossy: that is, they achieve high fast compression rates by losing information.
  • the loss is imperceptible: for example, with still images compressed using JPEG compression, the lost information is typically below the level of perception of the human eye.
  • JPEG and MPEG compression is useless. Accordingly, a need remains for a way to compress digital images that addresses these and other problems associated with the prior art.
  • FIG. IA shows a computer system designed to compress an image file according to an embodiment of the invention.
  • FIG. IB shows a computer system designed to decompress an image file according to an embodiment of the invention.
  • FIG. 1C shows two computers as shown in FIGs. IA and IB connected via a network for transferring a compressed image file.
  • FIG. 2 shows a close-up of an image file being compressed by the computer system of
  • FIG. IA is a diagrammatic representation of FIG. IA.
  • FIG. 3 shows five probability models used to compress the image of FIG. 2 on the computer system of FIG. IA.
  • FIG. 4A shows the image file of FIG. 2 being divided into boxes for compression on the computer system of FIG. IA.
  • FIG. 4B shows an image file divided into two different tessellations on the computer system of FIG. IA for compression.
  • FIGs. 5A and 5B show an animated image file being analyzed to determine a distance frame to compress the animated image file on the computer system of FIG. IA.
  • FIGs. 6A and 6B show the procedure used to compress the image file of FIG. 2 on the computer system of FIG. IA.
  • FIG. 7 shows the procedure used to divide the image of FIG. 2 into boxes for compression on the computer system of FIG. IA.
  • FIG. 8 shows a structure for an image file compressed according to the preferred embodiment of the invention.
  • FIGs. 9 A and 9B show the procedure used to decompress the image file of FIG. 2 on the computer system of FIG. IA.
  • FIG. 10 shows a Huffman coding tree according to the prior art.
  • Appendix A shows an implementation of the compression algorithm described herein, implemented in the C programming language and including comments.
  • FIG. IA shows a computer system 105 in accordance with the invention.
  • Computer system 105 includes a computer 110, a monitor 115, a keyboard 120, and a mouse 125.
  • Computer 110 includes hardware components, such as a processor 105, a memory 130, and a branch prediction apparatus (not shown).
  • Computer system 105 may also include other equipment not shown in FIG. IA, for example, other input/output equipment or a printer.
  • Computer system 105 stores image 130. Typically, image 130 is loaded into the memory of computer system 105.
  • Image 130 includes color palette 135, which specifies the colors used to display image 130.
  • Software 140 is installed in computer system 105.
  • Software 140 includes probability set 145, update module 150, compressor 155, distance frame generator 160, block locator 165, and size estimator 170.
  • Probability set 145 includes at least one model, and can possibly include multiple models, which can be used to predict the likelihood of the next pixel having a color that matches its left or upper neighbors.
  • Update module 150 is responsible for updating probability set 145 based on the actual colors of the pixel and its left and upper neighbors. Update module 150 can also select a different model from probability set 145, if needed. (Probability set 145 and update module 150 will be discussed further with reference to FIG. 3, below.)
  • Compressor 155 is responsible for encoding/decoding the color of the current pixel.
  • Distance frame generator 160 is responsible for generating distance frames between two frames of an animated image file. (Distance frame generator 160 is not used when an image file is static.)
  • Block locator 165 is responsible for locating blocks within image 130 that are different from the background color of the image.
  • Size estimator 170 is responsible for estimating the size of compressed image file 175 using different blocks located by block locator 165.
  • Compressor 155 uses the size estimations calculated by size estimator 170 to select the tessellation of the image (or the image frame) that will produce the smallest file size. Compressor 155 then uses compresses each block in the tessellation with the smallest file size for the image (or image frame). This compression uses the probability values in probability set 145.
  • computer system 105 includes decompressor 180.
  • Decompressor 180 is responsible for decompressing compressed image file 175. Since the contents of compressed image file 175 contain all the data of the image, decompressor 180 can completely reconstruct original image file 130 from compressed image file 175.
  • FIG. 1C shows two computers connected via a network for transferring a compressed image file.
  • computer system 105 stores a compressed image file.
  • Other computer systems such as computer systems 185 A, 185B, and 185C, are like computer system 105 and can make requests for the image file from computer system 105.
  • the compressed image file is then transferred over network 190 to the requesting computer, which can then decompress compressed image file 175 to reproduce original image file 130.
  • LANs local area networks
  • WANs wide-area networks
  • wireless networks and so on.
  • FIG. 2 shows a close-up of image file 130 being compressed by the computer system of FIG. IA.
  • image 130 includes a rectangular array of pixels.
  • Close-up section 205 shows some of the pixels in image 130 more closely.
  • pixel 210 has left neighbor 215 and upper neighbor 220.
  • Left neighbor 215 and upper neighbor 220 have the same color (represented by the cross-hatch pattern), and pixel 210 has the same color.
  • pixel 225 has the same color as its upper neighbor 230, but a different color than its left neighbor 210.
  • Pixel 235 has a color different than both its left neighbor 225 and its upper neighbor 240.
  • a person skilled in the art will recognize other combinations of colors for a given pixel and its respective left and upper neighbors.
  • FIG. 3 shows why it matters what the colors of the current pixel and its left and upper neighbors are.
  • FIG. 3 shows a probability set which includes five probability models used to compress the image of FIG. 2 on the computer system of FIG. IA.
  • Each of the probabilities 305-1, 305-2, 305-3, 305-4, and 305-5 reflects a combination of the colors of the current pixel and its left and upper neighbors; the probabilities are determined by the ratio of the individual probability relative to the sum of all probabilities with the same color combinations for the left and upper neighbors.
  • probabilities 305-1 and 305-2 are determined relative to all pixels whose left and upper neighbors have the same color
  • probabilities 305-3, 305-4, and 305-5 are determined relative to all pixels whose left and upper neighbors have different colors.
  • Probability 305-1 reflects the probability that the current pixel has the same color as both its left and upper neighbors (10/11).
  • Probability 305-2 reflects the probability that the current pixel has a different color than the left and upper neighbors, which have the same color (1/11).
  • Probability 305-3 reflects the probability that the current pixel has the same color as its left neighbor but a different color from its upper neighbor (10/21).
  • Probability 305-4 reflects the probability that the current pixel has the same color as its upper neighbor but a different color from its left neighbor (10/21).
  • Probability 305-5 reflects the probability that the current pixel has a different color than either its left or upper neighbors, which have different colors (1/21).
  • probabilities 305-1, 305-2, 305-3, 305-4, and 305-5 are the initial probabilities for each color combination, and are updated as the image is compressed.
  • probabilities 305-1 and 305-2 define the probabilities that the current pixel has the same color as its left and upper neighbors given that the left and upper neighbors have the same color. For example, given that the left and upper neighbors have the same color, it is initially assumed that the current pixel is ten times as likely as not to have the same color as its left and upper neighbors. As the model is updated, this probability changes. It is important to distinguish between the terms "probability set,” “probability model,” and “probability.” "Probability set” refers to the set of all probability models used in the compression.
  • probability model refers to the set of probabilities (which can also be called probability values) used in a single probability model. Typically, each probability model will include probabilities for similar conditions. “Probability” refers to the individual probability of a given condition happening in a single probability model. So, in FIG. 3, probability set 145 includes five probability models 305, 310, 315, 320, and 325, and each probability model (such as probability model 305) includes five probabilities 305-1, 305-2, 305-3, 305-4, and 305-5, shown as initial relative values 10, 1, 10, 10, and 1, respectively.
  • the different probability models allow for additional history to be used in predicting the current pixel's color.
  • the behavior of the next pixel will be similar to the behavior of other pixels that have similar color matches between the current pixel and its left and upper neighbors.
  • the current pixel has the same color as its upper neighbor but a different color from its left neighbor, it is expected that the next pixel will be colored relatively similarly to other pixels whose previous pixel has the same color as its upper neighbor but a different color from its left neighbor.
  • the colors of the left and upper neighbors select which probability to use within a probability model, the colors of the left and upper neighbors can be used to select the next probability set to use.
  • probability set 305 can be used in determining the probability of the next pixel's color.
  • probability set 320 can be used in determining the probability of the next pixel's color.
  • one model is used after one of the five probabilities occurs: that is, probability model 305 is used after the current pixel has the same color as both its left and upper neighbors, probability model 310 is used after the left and upper neighbors have the same color, but the current pixel has a different color, probability model 315 is used after the left and upper neighbors have different colors, and the current pixel has the same color as its left neighbor, probability model 320 is used after the left and upper neighbors have different colors, and the current pixel has the same color as its upper neighbor, and probability model 325 is used after the left and upper neighbors have different colors, and the current pixel has a different color than either its left or upper neighbor.
  • Update module 150 from FIG. 1 updates the individual probabilities of the probability models in the probability set. For example, assume that probability model 305 is currently being used, and the current pixel has the same color as its left neighbor, but not its upper neighbor. In this case, update module 150 updates probability 305-3. Update module 150 can also select the probability model to use with the next pixel: in this case, probability model 315.
  • FIG. 4A shows the image file of FIG. 2 being divided into boxes for compression on the computer system of FIG. IA.
  • image 130 includes two curved shapes and a rectangle.
  • image 130 is scanned to determine a background color. This is done by analyzing the border of the image. If there is a color that predominates the border of the image, this color is selected as the background color. Then, to divide image 130 into boxes, the image is scanned to determine vertical stripes in which pixels appear that differ from the background color. Two such stripes are present in FIG. 4A: stripe 405 and stripe 410. Each stripe is then divided into boxes, such that each horizontal row of pixels in the box includes at least one pixel with a color different from the background color. This forms boxes 415 and 420. Boxes 415 and 420 can then be compressed individually, with the remaining pixels in image 130 colored with the background color.
  • boxes are located by analyzing the image for stripes both horizontally and vertically.
  • the size of the compressed file is estimated using both horizontal and vertical striping, and the striping that produces the smaller file size is selected. This comparison is discussed further below with reference to FIG. 4B.
  • FIG. 4B shows an image divided into two different tessellations on the computer system of FIG. IA for compression.
  • image 425 includes one L-shaped object 430.
  • object 430 can be thought of as an approximation of the elliptical objects in box 415 of FIG. 4A.
  • object 430 could be enclosed by box 432, there are other ways to divide image 425.
  • image 425 can be tessellated into rectangular boxes, and each box compressed separately.
  • Two different tessellations are shown in images 435 and 440, respectively showing the boxes found when the first stripes are located vertically and horizontally.
  • boxes 445-1 and 445-2, or boxes 450-1 and 450-2 can be compressed according to the preferred embodiment of the invention.
  • the tessellations shown in images 435 and 440 may result in a smaller file size for the compressed image file. For example, if object 430 includes only one color (ignore for the moment dot 455), compressing box 432 would include two colors, and compression according to the preferred embodiment of the invention would be necessary.
  • boxes 445-1 and 445-2, or boxes 450-1 and 450-2 would each include only one color, and thus can be specified with only a location, size, and color.
  • dot 455 Although when dot 455 is ignored the two tessellations would result in compressed files of identical size, including dot 455 can make a difference.
  • the box that includes dot 455 is compressed using the preferred embodiment of the invention, rather than just storing a location, size, and color.
  • image 435 has an advantage, since box 445-2, which includes dot 455, is smaller than box 450-2 in image 440, and hence box 445-2 would compress to a smaller file size.
  • FIGs. 5 A and 5B show an animated image file being analyzed to determine a distance frame to compress an animated image file on the computer system of FIG. IA.
  • image frames 130-1 and 130-2 are consecutive frames from the animated image file.
  • each frame can be analyzed as described above with reference to FIG. 4A and compressed, this approach might not result in the best possible compression.
  • Careful inspection will reveal that the only change between image frames 130-1 and 130-2 is that rectangle 505 has moved downward slightly, as indicated by arrow 507.
  • a distance frame can be computed. This difference is shown in FIG. 5B as distance frame 130-3.
  • Box 510 has changed from the color of rectangle 505 to the background color of the animated image
  • box 515 has changed from the background color to the color of rectangle 515.
  • compressing boxes 510 and 515 of distance frame 130-3 would require less space than compressing image frame 130-2.
  • distance frame 130-3 can be decoded, and only the pixels that have changed from image frame 130-1 need to be redisplayed.
  • the preferred embodiment analyzes each frame four different ways. First, as discussed above, the size of each compressed frame is estimated using horizontal and vertical striping. Then the distance frame between the current frame and the previous frame of the animated image is calculated, and size estimations are calculated using horizontal and vertical striping on the distance frame. FIGs.
  • FIG. 6 A and 6B show the procedure used to compress the image file of FIG. 2 on the computer system of FIG. IA.
  • the color of the current pixel is compared with the colors of the current pixel's left and upper neighbors.
  • the appropriate probability value (and, if necessary, the color of the current pixel) is encoded at the applicable one of steps 610, 615, 620, 625, or 630 (FIG. 6B).
  • the model is updated. If multiple models are being used, at step 640 a new model is selected. If there are pixels remaining to be compressed, the procedure returns to step 605 for another pixel. Otherwise, at step 645, the indices of the color palette entries are compressed according to the model(s).
  • the encoding performed at the applicable one of steps 610, 615, 620, 625, or 630 is performed using a range encoder.
  • the range encoder is a variation of an arithmetic coder: the difference between an arithmetic coder and a range coder are primarily technical in nature and the distinction need not be explained here.
  • Arithmetic coding operates by assigning each possible symbol a range of probability values between 0% and 100%. Initially, the arithmetic coding covers the entire range from 0% to 100%. Then, as an individual symbol is encountered, the range is narrowed to include only the probabilities covered by that individual symbol. The process is then repeated, applied to the sub-range established by the previous symbol.
  • each range is defined to be exclusive of its upper limit, and so the value 0.1 is assigned only to the letter "B.”
  • a corollary of this definition is that the value 1.0 is excluded.
  • the range encoder When the range encoder encounters the first letter ("D"), it narrows the range of acceptable encodings from 0.0 through 1.0 to 0.6 through 1.0 (since the letter "D” is assigned the range 0.6 to 1.0). Since the second letter is "C,” the range is further narrowed to 0.72 through 0.84. This range is calculated by multiplying the range for the new symbol ("C") by the size of the range calculated so far (0.4, which gives the relative range of 0.12 through 0.24), and adding the resulting values to the low end of the previous range (resulting in 0.72 through 0.84). When the next symbol ("D") is encountered, the range is narrowed to 0.792 through 0.84. Table 1 shows the range of acceptable encodings after each symbol in the message is encountered.
  • the final range produced is 0.7938134784 though 0.7938162432.
  • the entire message is represented.
  • Decoding is accomplished by reversing the process. Again, the probabilities of the symbols in the message are known in advance. Then, the number is examined. Since it falls between 0.6 and 1.0, the first character of the message is "D.” The low value for the range of the determined character (0.6) is then subtracted from the encoded value, resulting in the value 0.1938134784. This value is then divided by the size of the range for the determined character (0.4), which produces 0.484533696.
  • the reader may wonder how this compresses the message, since 10 characters were needed to represent the encoding. This is explained by the fact that there are very few characters in the message. A longer message would show that the encoded message is shorter than the original message.
  • each of the four symbols “A,” “B,” “C,” and “D” can be assigned the same probability of 25%. This can be accomplished in many ways: for example, it can be preliminarily assumed that each symbol was encountered once for purposes of defining the initial probabilities. Then, as the symbols are encountered, the counts for the symbols can be updated. So, after the first symbol (“D") is encountered, its probability value can be updated to 40% (2 out of 5), with each of the other symbols reduced to 20% (1 out of 5).
  • the probability values for symbols “A” and “B” reduces to 16.67% (1 out of 6 for each), and the probability values for symbols “C” and “D” are changed to 33.33% (2 out of 6 for each). And so on.
  • Using dynamic probability values improves the speed of the encoding technique.
  • To compute the static probability values requires either selecting random probability values that may have no bearing on the actual message, or scanning the message to determine the probability values of the individual symbols in the message. Selecting random probability values can result in poor encoding; scanning the message requires performing two passes over the message (one to determine the probability values, and one to encode the message). With dynamic probability values, both disadvantages are avoided.
  • FIG. 7 shows the procedure used to divide the image of FIG. 2 into boxes for compression on the computer system of FIG. IA.
  • a background color is determined for the image. This is usually done by analyzing the colors of the pixels on the border of the image.
  • the image is divided into blocks. As described above with reference to FIG. 4A, in the preferred embodiment rectangular blocks are used. But any tessellation (division of the plane) can be used, provided that no pixel is included in more than one block, and that each pixel with a color different from the background color is included in a block.
  • Steps 715 and 720 are applicable only for animated images that include multiple frames.
  • Step 715 computes the distance frame between two frames in the animated image
  • step 720 computes tessellations of the distance frame.
  • multiple tessellations can be considered to find a tessellation that results in the smallest compressed image size.
  • step 725 the tessellation that results in the smallest compressed file size is selected, and then each block in the tessellation is compressed. Note that only the blocks need to be compressed: the remaining pixels include only the background color, which can be very easily identified.
  • FIG. 8 shows a structure for compressed image file 175 according to the preferred embodiment of the invention. In FIG.
  • the structure used for storing and for transmitting compressed image file 175 begins with preliminary information 805.
  • Preliminary information 805 is sent once per image, and specifies the size of the image (typically in two pixel dimensions: for example, 640x480), the color palette (typically as a table or a list of indices to color values), and the background color of the image (typically as an index into the color palette).
  • block information 810 specifies the location, size, and contents of a particular block in the compressed image file. The contents of the block are the compressed indices into the color palette for each pixel in the block. (In the preferred embodiment, pixels in each block are examined row by row, from the top of the block to the bottom, and within each row from left to right.
  • block information 810 can specify the location, size, and contents of box 415. Block information 810 can be repeated if there are multiple blocks in the image.
  • frame information 815 can be provided.
  • Frame information 815 specifies the type of frame (for example, the frame can be completely redrawn without reference to the prior frame, or the frame can be a distance frame, as described above) and the information about the frame.
  • the information about the frame can include one or more blocks, as described above. Preliminary information need not be resent until the next image. However, a person skilled in the art will recognize that some preliminary information (for example, a new color palette) can be sent, if desired.
  • compressed image file 175 In nowhere in FIG. 8 is the structure of compressed image file 175 described as including the model(s) used in compression. The reason the models are not included in compressed image file 175 is that the models do not need to be stored. Instead, the models can be reconstructed as compressed image file 175 is read. For example, if when the image is compressed a particular pixel has the same color as its left and upper neighbors, the index into the color palette stored in the compressed image file will reflect this. Thus, the models can be reconstructed when compressed image file 175 is read, and do not need to be stored in compressed image file 175. This is an advantage over Huffman codes, which must be stored with the image file.
  • FIGs. 9A and 9B show the procedure used to decompress compressed image file 175 of FIG. 2 on the computer system of FIG. IA.
  • step 905 the probability value for the current pixel is decoded.
  • the probability value is analyzed to determine whether the color of the current pixel is supposed to be the same as either the left or upper neighbor of the current pixel. If the current pixel is the same color as either the left or upper neighbor of the current pixel, then at step 915 the color of the current pixel is copied from the left or upper neighbor, as appropriate, of the current pixel. Otherwise, at step 920, the color of the current pixel is decoded from the compressed image file.
  • step 925 FIG.
  • step 9B the probability model is updated, and if necessary, at step 930 a new probability model is selected. Then, if more pixels remain to be decoded, the process returns to step 905. Otherwise, at step 935, the decompressed image is displayed.
  • step 930 as discussed above with reference to FIG. 3, the use of different probability models helps to improve the compression by focusing each probability model on a different probability value. For example, consider probability model 305, and assume that the current pixel and its left and upper neighbors have the same color. Because the current pixel had the same color as its left and upper neighbors, the compression technique expects that the next pixel will have the same color as its left and upper neighbors.
  • probability value 305-1 will be further increased, as opposed to any of probability values 305-2, 305-3, 305-4, and 305-5.
  • the bigger the available range the fewer the number of bits necessary to compress the message.
  • the probability models are able to compress the image using fewer bits, resulting in better compression than would otherwise occur.
  • the decompressor is able to decompress compressed image file 175 without reading the probability models from compressed image file 175 because the decompressor is able to recreate the probability models on the fly.
  • the decompressor starts with initial probability models, the same as shown in FIG. 3.
  • the compressor/decompressor resembles Lempel-Ziv compression, hi the Lempel-Ziv compression, the compressor and decompressor build dictionaries as the file is read for compression/decompression.
  • the instant invention does not use a dictionary, the probability models can be built by both the compressor and decompressor as they read the image file and the compressed file, respectively.
  • Appendix A shows an implementation of the compression algorithm described herein.
  • the comments describe a preferred embodiment of the algorithm implemented in the source code.
  • the comments in Appendix A can be thought of as pseudo ⁇ code.
  • the routine ac_out is a standard range encoder.
  • the routine UpdateModel * is responsible for updating the model to reflect the new probability
  • * pi is the index into the color palette for the left neighbor pixel.
  • * plu is the index into the color palette for the upper neighbor pixel.
  • * ptr is a two-dimensional array storing the indices into the color
  • * xl marks the low end of the range of probability values for a color
  • * xh marks the high end of the range of probability values for a color * in the color palette.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Of Band Width Or Redundancy In Fax (AREA)
  • Image Processing (AREA)

Abstract

Pour comprimer une image, on considère chacun de ses pixels dont la couleur, représentée par un indice de la palette de couleurs, est comparée à celle des pixels voisins de gauche et du dessus, puis codée dans une entrée de la palette de couleurs. On actualise la loi de distribution en fonction de la couleur du pixel et de celle de ses voisins de gauche et du dessus. Un fois tous les pixels codés, la palette de couleurs est optimisée à l'aide de la loi de distribution, et on comprime les indices de la palette de couleurs des pixels. Dans l'exécution préférée la compression se fait en une seule passe sur les pixels de l'image et la loi de probabilité est actualisée dynamiquement lors de la compression de chaque pixel.
PCT/US2004/027389 2004-08-23 2004-08-23 Compression d'images utilisable avec des images animees WO2006022741A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/US2004/027389 WO2006022741A1 (fr) 2004-08-23 2004-08-23 Compression d'images utilisable avec des images animees

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2004/027389 WO2006022741A1 (fr) 2004-08-23 2004-08-23 Compression d'images utilisable avec des images animees

Publications (1)

Publication Number Publication Date
WO2006022741A1 true WO2006022741A1 (fr) 2006-03-02

Family

ID=35967800

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2004/027389 WO2006022741A1 (fr) 2004-08-23 2004-08-23 Compression d'images utilisable avec des images animees

Country Status (1)

Country Link
WO (1) WO2006022741A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2521495A (en) * 2013-12-20 2015-06-24 Canon Kk Method and apparatus for transition encoding in video coding and decoding
JP2016529783A (ja) * 2013-07-12 2016-09-23 クゥアルコム・インコーポレイテッドＱｕａｌｃｏｍｍ Ｉｎｃｏｒｐｏｒａｔｅｄ パレットベースのビデオコーディングにおけるパレット予測
CN110225345A (zh) * 2013-12-27 2019-09-10 寰发股份有限公司 用于主颜色索引图编码的方法及装置
US11259020B2 (en) 2013-04-05 2022-02-22 Qualcomm Incorporated Determining palettes in palette-based video coding

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6038346A (en) * 1998-01-29 2000-03-14 Seiko Espoo Corporation Runs of adaptive pixel patterns (RAPP) for lossless image compression
US6118552A (en) * 1993-09-21 2000-09-12 Fuji Xerox Co., Ltd. Image signal encoding apparatus
US6512793B1 (en) * 1998-04-28 2003-01-28 Canon Kabushiki Kaisha Data processing apparatus and method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6118552A (en) * 1993-09-21 2000-09-12 Fuji Xerox Co., Ltd. Image signal encoding apparatus
US6038346A (en) * 1998-01-29 2000-03-14 Seiko Espoo Corporation Runs of adaptive pixel patterns (RAPP) for lossless image compression
US6512793B1 (en) * 1998-04-28 2003-01-28 Canon Kabushiki Kaisha Data processing apparatus and method

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11259020B2 (en) 2013-04-05 2022-02-22 Qualcomm Incorporated Determining palettes in palette-based video coding
JP2016529783A (ja) * 2013-07-12 2016-09-23 クゥアルコム・インコーポレイテッドＱｕａｌｃｏｍｍ Ｉｎｃｏｒｐｏｒａｔｅｄ パレットベースのビデオコーディングにおけるパレット予測
GB2521495A (en) * 2013-12-20 2015-06-24 Canon Kk Method and apparatus for transition encoding in video coding and decoding
GB2521606A (en) * 2013-12-20 2015-07-01 Canon Kk Method and apparatus for transition encoding in video coding and decoding
GB2521684A (en) * 2013-12-20 2015-07-01 Canon Kk Method and apparatus for transition encoding in video coding and decoding
GB2523301A (en) * 2013-12-20 2015-08-26 Canon Kk Method and apparatus for transition encoding in video coding and decoding
GB2521495B (en) * 2013-12-20 2016-05-18 Canon Kk Method and apparatus for transition encoding in video coding and decoding
US9516342B2 (en) 2013-12-20 2016-12-06 Canon Kabushiki Kaisha Method and apparatus for transition encoding in video coding and decoding
CN110225345A (zh) * 2013-12-27 2019-09-10 寰发股份有限公司 用于主颜色索引图编码的方法及装置
CN110225345B (zh) * 2013-12-27 2022-07-19 寰发股份有限公司 用于主颜色索引图编码的方法及装置

Similar Documents

Publication Publication Date Title
US7295704B2 (en) Image compression usable with animated images
US11252441B2 (en) Hierarchical point cloud compression
CN113557745B (zh) 点云几何结构填充
Sneyers et al. FLIF: Free lossless image format based on MANIAC compression
JP2968582B2 (ja) デジタルデータを処理するための方法および装置
US20180189982A1 (en) Attribute mapping to encode and decode 3d models
US5303313A (en) Method and apparatus for compression of images
EP1431919B1 (fr) Procédé et appareil pour le codage et le décodage d'objets tridimensionnels au moyen d'octrees
US6243416B1 (en) Image coding and decoding methods, image coder, and image decoder
Motta et al. Lossless image coding via adaptive linear prediction and classification
US20030123545A1 (en) Segment-based encoding system using segment hierarchies
US20040114819A1 (en) Image processing system for compressing image data including binary image data and continuous tone image data by a sub-band transform method with a high-compression rate
US7082166B2 (en) Decoder for decoding segment-based encoding of video data using segmentation performed at a decoder
US7551773B2 (en) Encoding image data using mask lookup table and decoding such encoded image data
JP3872217B2 (ja) ディザ画像の２値表現処理方法、ディザ画像の圧縮２値表現圧縮解除方法、及びディザ画像の圧縮及び圧縮解除システム
JP5101962B2 (ja) 画像符号化装置及びその制御方法並びにコンピュータプログラム
WO2006022741A1 (fr) Compression d'images utilisable avec des images animees
US7103103B2 (en) Segmenting encoding system encoding video data using segment-by-segment kinetic data including change information in addition to translation information
EP1087533A1 (fr) Procede et appareil permettant de constituer une table de codage, dispositif de quantification de vecteur, dispositif et procede de compression de donnees, dispositif et procede de decompression de donnees, systeme de compression/decompression de donnees
Pensiri et al. A lossless image compression algorithm using predictive coding based on quantized colors
Sayood Data Compression.
JP3646074B2 (ja) 情報埋込装置及び情報抽出装置
JP4743884B2 (ja) 画像符号化装置及びその制御方法
JP2010074444A (ja) 画像圧縮装置、画像圧縮方法、画像復号装置、画像復号方法、及び、コンピュータプログラム
JP4743883B2 (ja) 画像符号化装置およびその制御方法

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase