Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T9/00—Image coding
  • G06T9/005—Statistical coding, e.g. Huffman, run length coding

Definitions

• the present invention relates generally to data reduction techniques for digitized images and, more particularly, to data reduction utilizing different image representation techniques for different parts of the image.
• Digital imaging systems have improved the process of creating, editing and rendering images. In particular, such systems have decreased the amount of processing time necessary to produce a final product. Nonetheless, the ability of digital imaging systems to process and, particularly, to render images is limited by the memory capacity of the systems.
• an image is often divided into fixed spatial coordinates on a rectangular grid where each sample point has one color and is referred to as a picture element, commonly known as a pixel or a "dot" (not to be confused with the halftone dot used in the printing industry).
• a picture element commonly known as a pixel or a "dot" (not to be confused with the halftone dot used in the printing industry).
• Such an image is usually referred to as a raster image and is typically represented and stored in a format which uses one or more bits per pixel to identify the color of each pixel.
• the total amount of data necessary to represent an image depends on several factors, some of which include the image size, the resolution of the image, and the number of bits per pixel.
• Raster Image Processor would typically not be able to handle the volume of data in a printing format of a 1200 dot per inch (dpi) image file represented in contone raster format, such as what might be required for imaging data for a map.
• a standard printing form sized 32 inches by 44 inches full of 1200 dpi, 8 bits-per-pixel images would require about 2 gigabytes of memory, well beyond that available on most rendering/RIP workstations.
• the image consists of some photographic content and a large portion of "line work", i.e., text or geometric objects which delineate areas of constant color which are amenable to representation with a small number of bits per pixel.
• the present invention provides a system and method of reducing the volume of data required to represent an image by "tiling" the image (i.e., subdividing the image into smaller, preferably rectangular areas) and representing each tile using a "fill” operation and/or one or more 1 -bit-per-pixel masks, or by representing each tile with an indexed image when such a representation of the tile requires fewer bits per pixel than in the original image.
• a digital image described by raster content encoding the pixels using an original number of bits per pixel is initially provided.
• the image is then divided into smaller sub-areas or tiles.
• Each of the tiles is individually analyzed to determine whether a data volume savings is worthwhile if the techniques of the present invention is utilized. In particular, this is accomplished by comparing the required number of masks necessary to represent the tile to a threshold number of masks, typically set equal to the original number of bits per pixel representing the image or predetermined to be some smaller number deemed to result in a worthwhile data volume savings.
• the threshold number of masks can be predetermined or user-defined via a user input or a system configuration setting.
• a representation for the tile using fills and/or masks is generated. Otherwise, a representation using the original number of bits per pixel or using an indexed image requiring fewer bits per pixel for the tile is used.
• a representation of an image may specify a "default" color (typically white but also commonly colorless) so that any pixel which has that color does not need to be represented explicitly.
• a default color typically white but also commonly colorless
• a test is made to determine whether there are no default-colored pixels in the tile (a determination based on the image) and whether a non-default background color should be chosen (a user option).
• the default is selected as the background color.
• a non-default color is selected as the background color and a representation is generated to fill the tile with the selected non-default background color. Then, if there are any non-background colors, for each such color, a representation is generated for the pixels in an area of the tile using a 1-bit-per-pixel mask.
• the area of the mask can be defined by the boundaries of the tile or, alternatively, it can be a minimally sized area within the tile in which pixels of the current color are located.
• the test could query whether there are only a few default-colored pixels in the tile. If there are only a few and default-colored pixels can be explicitly represented (a function of the representation system or conditions established prior to rendering the image), then a non-default background is selected and a representation is generated to fill the tile with the selected non-default background color as described above. However, if there are a large number of default-colored pixels in the tile, the default is selected as the background color.
• the test could simply query whether there are any default-colored pixels in the tile without testing whether 5 a non-default background should be chosen. In this case, if there are no default-colored pixels in the tile, a non-default background will be selected. If there are any default-colored pixels, a default background is selected. In yet another embodiment, a background color is simply selected and, if it needs to be rendered, instructions to fill the tile with that color are generated. In still another further
• a background color is not chosen at all and the tile is represented using masks alone, if a data savings was deemed achievable by use of such masks.
• the required number of masks to represent the tile is greater than or equal to the threshold number
• a representation is generated for the tile using the original number of bits per pixel provided by the original raster content.
• the original format and number of bits per pixel are used to represent the entire tile.
• the original format and number of bits per pixel can be used only in a minimal area within the tile. In this case, the tile is examined to
• the tile may be represented such that the value of each pixel is encoded as an index which references
• a table of the colors in the tile (as distinct from the entire image) as will be appreciated by those skilled in the art and others. For example, if there are 16 or fewer colors in the tile then a 4-bit index is sufficient to identify the color so that only 4 bits per pixel are required for the tile. In this embodiment, a test can be performed to indicate whether a data savings can be achieved using the index as compared to using the original bits per pixel for the tile. If so, an index is used to represent the colors in the tile. If not, the original bits per pixel are used.
• a technique of indexing tiles in the image could be used.
• the original image is first divided into tiles. Then, for each tile, the colors are identified and a test is performed to determine whether a data savings can be achieved using an indexed representation for the tile. If so, the tile is represented using an index. If not, the tile is represented using the original bits per pixel.
• FIGURE 1 is a block diagram of a system in which the present invention is implemented
• FIGURE 2 is a flow diagram illustrating the logic of the present invention
• FIGURE 3 is a flow diagram illustrating how the tiles are represented using fills and/or masks
• FIGURE 4 is a flow diagram illustrating alternative logic for representing a tile with its original number of bits per pixel or a smaller index
• FIGURE 5 is a flow diagram illustrating yet further alternative logic for representing a tile with its original number of bits per pixel or a smaller index
• FIGURE 6 is a flow diagram illustrating alternative logic to that shown in FIGURE 2.
• FIGURE 1 is a simplified block diagram of a system 20 in which the present invention is implemented.
• the system 20 includes a processing unit 24 that receives and processes an original digital image 22 in accordance with the present invention.
• the original image 22 is created using methods well known in the art.
• the original image 22 is stored in a computer-readable file having a file format for representing bit-mapped graphic data either directly or in an indexed manner. Examples of such file formats include tagged image file format (TIFF), TIFF for Information Technology (TIFF/IT), Scitex (Handshake or native) Line Work (LW), and others.
• TIFF tagged image file format
• TIFF/IT TIFF for Information Technology
• Scitex Handshake or native Line Work
• LW Line Work
• the original digital image includes a plurality of pixels. Each of the pixels in the original image is represented by an original number of data bits, which may be an index, necessary for describing the characteristics of the pixel, such as color.
• the original image 22 is transmitted to the processing unit 24 from any data source including, but not limited to, a local data source such as a computer workstation or server directly connected to the processing unit, a remote data source, or a computer-readable medium storing the original image such as a floppy disk.
• the processing unit 24 includes an input/output port for receiving the transmitted original image 22.
• the original image 22 can simply be stored on and retrieved from memory of the processing unit 24.
• the processing unit 24 also includes a memory 23 for storing data and instructions necessary for reducing the volume of data representing the original digital image and a processor 25 for executing the instructions. More specifically, the memory 23 stores a data reduction program 27 formed in accordance with the present invention for reducing the volume of data representing the original image by using tiling, masking and/or indexing techniques. Once the processor 25 executes the data reduction program 27, a reduced-data-volume digital image 26 is produced. The reduced-data-volume digital image 26 can then be transmitted to another device for further processing or may be further processed by the processing unit 24.
• the processing unit 24 can be a general purpose computer workstation or can be located within a Raster Image Processor (RIP) or any image rendering device. Accordingly, the reduced data volume digital image 26 may be directly transferred to other processes typically implemented by any of these devices.
• RIP Raster Image Processor
• the data reduction program 27 reduces the volume of data necessary to represent an image, in a format such as PostScript or Portable Document Format (PDF) which accommodates a mixture of raster formats, by using tiling, masking, and/or indexing techniques.
• PDF Portable Document Format
• the image is divided into smaller sub-areas or tiles.
• Each of the tiles is then individually analyzed to determine whether a data volume savings is possible and worthwhile. This is essentially accomplished by comparing the number of bits per pixel required to represent the tile if the data reduction methods of the present invention are utilized with the original number of bits per pixel describing the image or some other threshold. If a data volume savings is possible and worthwhile, the data reduction methods of the present invention are used.
• a background color for the tile is chosen and a representation for filling the tile with the background color is provided.
• a representation is generated for identifying any remaining non-background colors in the tile using masks or an indexed representation.
• the "color" of a pixel is not necessarily restricted just to an intensity for each component of a color space such as red-green-blue (RGB) or cyan-magenta-yellow-black (CMYK), but can also include degrees of transparency or other properties.
• RGB red-green-blue
• CYK cyan-magenta-yellow-black
• a representation of an image may specify a "default" color (typically white but also commonly colorless) so that any pixel which has that color does not need to be represented explicitly.
• a default color typically white but also commonly colorless
• a given tile may be designated with a "background” color, which is typically the dominant color in the tile.
• a mask is a 1-bit-per-pixel raster image. A mask is applied to an area of an image by first selecting a color and then, for each pixel of the mask which has the value 1, the corresponding pixel of the image area to which the mask is applied is changed to the selected color. For each pixel of the mask which has value 0, the corresponding pixel of the image area is left unchanged.
• FIGURE 2 is a flow diagram that illustrates the logic of the data reduction program 27 of the present invention as implemented by the processing unit 24.
• an image is divided into tiles or blocks such that each tile represents a portion of the overall image. This can be accomplished in a predetermined fashion or in a dynamic fashion based on the tile content.
• the tiles may be uniformly sized. However, it will be appreciated by those skilled in the art and others that the tile size can be varied from tile to tile.
• the number of tiles into which the image is divided can also vary and depends primarily upon the number of colors represented in each tile. The purpose of dividing the image into tiles is to create smaller areas with fewer colors that, therefore, need less data to represent the image than required by the original raster content.
• a first tile is retrieved. Then, the colors in the tile are identified.
• an array which includes, for each color in the total image, a counter for the number of pixels for that color as well as elements for storing the minimum and maximum horizontal and vertical coordinates of pixels of that color in the tile.
• Each pixel in the tile is analyzed such that every time a pixel of a particular color is found, the counter and minimum and maximum coordinates for that particular color are updated.
• the counters are initially set to a "0" value, if a counter has a positive value after analyzing the pixels in the tile, then it is known that the color represented by such counter is present in the tile.
• the colors in each tile are tracked.
• the colors in the tile can be identified by setting a flag to indicate merely the presence of a particular color in the tile.
• a sorted list can be maintained. Once the number of colors in the list equals or exceeds the threshold number of masks, the generation of the list may be aborted and the tile represented using the original format and number of bits per pixel or using an index for the tile (as distinct from the entire image). If this embodiment is implemented, the alternative embodiments (described below) that permit the identification of minimally sized areas within the tile can only be implemented if the tile is completely analyzed, but require a modification because complete information about all the colors in the tile and the maximum and minimum horizontal and vertical coordinates for each color in the tile need to be identified for those embodiments.
• a test is made to determine whether the required number of masks is greater than or equal to some threshold number of masks.
• the required number of 1-bit masks to represent the image is logically determined by 1) the colors of the pixels in the tile together with 2) a decision by the user of the system, or predetermined as a fixed condition of the system, as to whether or not to use fills for background colors when possible.
• n denote the number of colors in the tile; let t be 0 if there is no default color or, if there is a default, there are no default-colored pixels in the tile; let t be 1 if there is a default color and there are default-colored pixels in the tile; and let f be 0 if fill operations are not to be used for background colors and 1 if they are.
• the required number of masks for a tile is n-(t
• the threshold number of masks might be set equal to the original number of bits per pixel required to represent the entire image as provided in the original raster content. Essentially, a determination is being made at a step 38 regarding whether a data volume savings can be made for this tile using the masks. However, the threshold number of masks is preferably set to some number less than the original number of bits per pixel either to satisfy a certain level of data volume savings or if an indexed representation would be more effective. For example, a user could decide that although the tile is originally represented by 8 bits per pixel, a satisfactory data reduction would involve using no more than 4 bits per pixel to represent the tile, i.e. 4 masks. Thus, the user would select the threshold number of masks to equal 5. It will be appreciated by those skilled in the art and others that the threshold number of masks can be hard-coded into the algorithm or user-defined as a user input or system configuration setting, or varied dynamically depending on the size and/or contents of the tile.
• a representation for the tile is generated using the original format and number of bits per pixel or using an index which uses fewer bits than the original number. See step 40. Generating such a representation by using either computer instructions, image description operators, or data structures is well-known in the art and thus will not be described in any further detail here. With respect to the option of indexing, if the number of colors in the tile were less than or equal to 16, then a 4-bit index would be sufficient to represent each pixel and would result in a data savings if it is less than the original number of bits per pixel. If the number of colors were less than or equal to 256, an 8-bit index would suffice. Indexed representations are well understood by those skilled in the art and others.
• step 40 or 46 After either step 40 or 46, another test is made at a step 42 to determine whether there is another tile to process. If there is another tile to process, the next tile is retrieved at a step 44 and the process described above is repeated as shown in FIGURE 2 using the next tile. If there are no other tiles to analyze, the process ends.
• FIGURE 3 is a flow diagram illustrating how tiles are represented using fills and or masks as required in step 46 of FIGURE 2.
• a two-part test is made to determine whether both of the following conditions have been met: 1.) whether there are no default-colored pixels in the tile and 2.) whether a non-default background for the tile should be chosen. If the answer to either condition is no, meaning that either the tile includes default-colored pixels or that a non-default background should not be chosen, the default color is selected as the background color for the tile at a step 62. Beyond the mere selection of the default as the background color for the tile, no data is necessary to represent the default background.
• a non-default background color is selected at a step 64.
• the choice of the background color could be arbitrary from among the colors in the tile. However, it will be appreciated by those skilled in the art and others that the choice of the background color could be based upon some criterion. For example, a mask for each of the colors in the tile could be calculated, as will be described in further detail below, and this data could then be compressed. The color of the mask that compresses the least could then be chosen as the background color.
• the color which has the greatest number of pixels in the tile, or for which the bounding rectangle is largest could be determined and selected as the background color. This determination could be made by examining the counters associated with the array of potential colors available for the tile. In any case, once the non-default background color is selected, a computer-readable representation is generated to fill the tile with the non-default background color at a step 66.
• step 60 is optional. Instead of performing the test indicated in step 60 (that is, removing step 60 entirely), the algorithm can be hard-coded to automatically select the default color as the background, or if the default color (e.g. white) is renderable, to select even some other color as the background color. Alternatively, the algorithm can simply be hard-coded to select any color identified in the tile as the background color either arbitrarily or based upon some criterion such as those described above with reference to step 64.
• the default color e.g. white
• the algorithm can simply be hard-coded to select any color identified in the tile as the background color either arbitrarily or based upon some criterion such as those described above with reference to step 64.
• steps 60-66 can be eliminated altogether such that there is no background color chosen and processing continues as described below such that the tile is represented using only masks if a data savings was deemed worthwhile.
• test could query whether there are only a few renderable default-colored pixels in the tile. If there are only a few, then a non-default background is selected and a representation is generated to fill the tile with the selected non-default background color as described above. However, if there are a large number of default-colored pixels in the tile, the default is selected as the background color.
• the test could simply query whether there are any default-colored pixels in the tile without testing whether a non-default background should be chosen. In this case, if there are no default-colored pixels in the tile, a non-default background will be selected. If there are any default-colored pixels, the default is selected as the background color.
• step 68 a test is made to determine whether there are any unprocessed non-background colors in the tile. See step 68. If the answer is no, processing returns to FIGURE 2. See step 70. In particular, if this happens at the first entry to block 68, then the tile is entirely the background color. In this case, if the tile is the default color, no data is necessary to represent the tile as described above and, thus, the tile most certainly requires far fewer bits per pixel than originally required for that area in the image.
• the tile is a solid non-default color
• bits necessary to identify the color and the tile boundaries and to fill that tile with the identified color are still far fewer than the original number of bits per pixel for that area of the image provided the tile is not too small.
• a data volume savings is achieved.
• a non-background color is identified at a step 72.
• a representation is generated for identifying the pixels in an area of the tile with the current color as a 1 -bit-per-pixel mask.
• the mask identifies which pixels within an identified boundary are to be rendered with the identified color.
• the area of the mask is defined by the boundaries of the tile and, thus, the mask applies to the entire tile.
• the area in the tile for which the mask is provided can be minimized by determining the minimal area in which the current color is located. This is accomplished by determining the minimum and maximum horizontal and vertical coordinates in the tile where the current color is located.
• the boundaries of the minimal area can be determined as those skilled in the art and others will appreciate.
• the mask for the current color can be provided in the identified minimal area only.
• you had a circle of diameter 200 pixels in a square tile of 500 pixels instead of storing a mask with a 1 -bit-per-pixel mask for 500 by 500 pixels you can store only the information for the smallest square containing the circle, i.e., 1 -bit-per-pixel for 200 by 200 pixels plus the information about the position of the mask within the larger square.
• processing returns to step 68 to determine whether there are any more unprocessed non-background colors and the process described above is repeated as shown in FIGURE 3 and as described above.
• a next non-background color is identified at step 72 and a mask is generated. This occurs until all non-background colors have been processed.
• FIGURE 4 illustrates how generating a representation for a tile using the original format and number of bits per pixel or using a smaller index, as required at step 40 of FIGURE 2, can alternatively be accomplished in accordance with another embodiment of the present invention.
• the tile instead of generating a representation providing the tile boundaries and the original number of bits per pixel or providing a smaller index for the area in the image corresponding to the designated tile area, the tile can be represented by identifying a background color for the entire tile, identifying boundaries for a minimal area in the tile in which the remaining colors are contained, and utilizing the original format and number of bits per pixel or a smaller index for the identified minimal area only.
• this corner area can be identified so that the original number of bits per pixel or smaller index need only be used for this minimal area.
• the rest of the tile is the default color, no further data need be provided.
• the rest of the tile is some other color, a representation providing the color and tile boundaries and requesting that the tile be filled with the identified color is also generated.
• FIGURE 5 illustrates another alternative embodiment for the method of generating a representation for a tile using the original format and number of bits per pixel or using a smaller index, as required at step 40 of FIGURE 2.
• a test is made to determine whether using an index to represent the colors in the tile results in a data savings over the original number of bits per pixel. If the answer to this test is yes, then the colors in the tile are represented using an index at a step 94. If the answer to this test is no, the tile is represented with the original bits per pixel at a step 92. Then, at block 96, processing returns to FIGURE 2.
• FIGURE 6 is a flow diagram illustrating alternative logic to that shown in
• FIGURE 2 In general, rather than determining whether the required number of masks is greater than or equal to the threshold number of masks, a test is made to determine whether a data savings can be achieved simply by using an index for the tile.
• the image is first divided into tiles at a step 100. Then at a step 102, a first tile is retrieved. Next, the colors in this tile are identified at a step 104, and a test is made at a step 106 to determine whether a data savings can be achieved by representing the tile using an index. If the answer to the test at step 106 is no, the original bits per pixel are used to represent the tile. See step 108.
• a data reduction method in accordance with the present invention has a number of advantages. By reducing the number of bits required to represent the image to. less than originally required, the file containing the image can be more easily stored and processed to render the image. In many cases, it makes rendering the image possible. In other cases, it speeds up the processing time.

Landscapes

• Engineering & Computer Science (AREA)
• Multimedia (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Image Processing (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=22595530

Family Applications (1)

Country Status (4)

Family Cites Families (3)

• 2000
  • 2000-11-09 AU AU12612/01A patent/AU1261201A/en not_active Abandoned
  • 2000-11-09 JP JP2001536998A patent/JP2003514416A/ja active Pending
  • 2000-11-09 EP EP00974209A patent/EP1238370A1/de not_active Withdrawn
  • 2000-11-09 WO PCT/CA2000/001307 patent/WO2001035340A1/en not_active Application Discontinuation

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events