JP3998115B2 - Image processing apparatus and image processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
  The present invention relates to color document image data, monochrome document image data, and other code compression methods, code compression apparatuses, and document images to which code data obtained by code compression of image data is further compressed. The present invention relates to an image processing apparatus such as a filing system.
[0002]
[Prior art]
  In an image processing apparatus, when a large amount of data such as a color digital image is stored in a storage device having a limited capacity, a data compression technique is often used. Furthermore, when the code amount when the image data is encoded by a certain compression method exceeds the capacity of the storage device, it is necessary to reduce the already encoded and stored data by that excess. This information reduction is usually an irreversible process using visual redundancy included in the image data.
[0003]
  Conventionally, in a device that compresses and stores images of a large number of pages in a storage device such as a hard disk (HDD), when the HDD capacity is full, the compressed data of the pages that have been stored so far is reduced. A page can be stored. Data reduction is achieved by reducing the image quality such as resolution and gradation and increasing the compression rate.
[0004]
  Hierarchical coding is an encoding method that can easily reduce information from such encoded data. In the case of non-hierarchical encoding, for example, JPEG baseline, it is necessary to go through the troublesome procedure of once decompressing the compressed code and recompressing it again at a different compression rate. However, with hierarchical encoding such as JPEG extended progressive, information can be reduced simply by discarding data in layers that are not important in terms of image quality.
[0005]
  On the other hand, there is a fixed-length compression method as a method for code-compressing color image data more easily than JPEG. This is because color image data is converted into block units, the lightness component is converted into average lightness, lightness dispersion, and each pixel quantized value by the GBTC (generalized block truncation coding) method. This is a method of generating and using this as a code. Since the generated code length is the same for each block (fixed length compression), the code is easy to handle. For example, since the transfer speed when transferring the code can be made constant, there are advantages such as simplified flow control and easy address calculation when directly processing and editing the stored code. For example, in Japanese Patent No. 2608284, the ratio of the code length related to lightness and the code length related to color is changed depending on whether or not an edge is further included in the block in such a fixed-length compression method. It is described to improve image quality.
[0006]
[Problems to be solved by the invention]
  In the above prior art, the hierarchical coding method such as JPEG extended progressive has a drawback that the processing is complicated. In addition, the method described in Japanese Patent No. 2608284 is a comparative process.TargetAlthough it is simple, there is a drawback that the information reduction (compression rate readjustment) from the state of the compression code is not easy due to the code structure, or that a new problem is caused by the reduction. For example, if the code relating to the color is deleted, the code lengths per block will vary, and the merit of the fixed length compression will be lost.
[0007]
  In view of the above-described problems of the prior art, the present invention can easily process and edit image data compression codes, can easily increase the compression ratio, and can reduce image quality deterioration at that time. An object is to provide an apparatus and an image processing apparatus to which the apparatus is applied.
[0008]
[Means for Solving the Problems]
  The present invention employs a fixed-length compression method that makes it easy to handle codes, and removes low-importance code portions from fixed-length encoded data consisting of a plurality of code portions obtained by code compression of image data ( (Delete) or reduced replacement. Further, if necessary, the code part having the next lowest importance in the encoded data of the compression result is removed, and the remaining significant code part is re-encoded.
[0009]
  For example, for each block of color image data, a first code representing the lightness average of the pixels in the block, a second code representing the lightness variance of the pixels in the block, a third code representing the individual lightness values in the block, Then, with respect to the encoded data that is code-compressed to the fourth code representing the color average of the pixels in the block, the encoded data is obtained by removing or reducing and replacing the fourth code part. If the image data is a normal business document image, the least important is the color code. Therefore, even if the color information is degenerated or lost, the meaning of the document can remain.
[0010]
  Furthermore, if necessary, the third code is reproduced according to the values of the first code and the second code with respect to the encoded data composed of the first code, the second code, and the third code of the compression result. The first code part and the second code part are removed, and only the regenerated third code is used. Specifically, when the lightness variance value of the second code is larger than a predetermined threshold value, the original third code is used as it is after the regeneration, and when it is smaller than the predetermined threshold value, the lightness average of the first code is used. In response to this, the third code is regenerated. For example, a block having a large brightness variance value is an area where characters are written, and it is unlikely that the meaning of the document is lost even with the third code alone. A block having a small lightness variance value is a region such as a photograph, for example, and image quality deterioration can be reduced by regenerating the lightness value of each pixel in the block according to the lightness average.
[0011]
  In the case of monochrome image data, the encoded data originally represents a first code representing the lightness average of the pixels within the block, a second code representing the lightness variance of the pixels within the block, and the lightness values of the individual pixels within the block. Since it is composed of the third code, when further compression is required, the encoded data consisting of the first code, the second code, and the third code is directly compressed as described above. Only the code of 3 may be used.
[0012]
  The image processing apparatus and the image processing method according to the present invention can be defined as follows.
[0013]
  Image processing according to claim 1 includes:
  Color conversion means for converting color image data into lightness component data and color difference component data;
Lightness encoding means for compressing the lightness component data in a predetermined block unit and generating codes of average lightness value, lightness variance value, and lightness quantization value;
Color difference encoding means for generating a code of an average color difference value by compressing the color difference component data at a fixed length in block units;
Accumulation means for accumulating encoded data composed of codes of the average lightness value, the lightness variance value, the lightness quantization value, and the average color difference value;
Color difference code removing means for removing the code of the average color difference value stored in the storage means in units of blocks;
Lightness code requantization means for removing the code of the lightness average value and lightness variance value stored in the storage means for each block;
It is characterized by having.
[0014]
  In addition to the configuration of the image processing device according to claim 1, the image processing device according to claim 2 includes:
  The lightness code requantization means includes a table in which the number of 1s in the code of the lightness quantization value and the average lightness value are associated with each other.
The lightness code requantization means reads the sign of the average lightness value, lightness variance value, and lightness quantization value accumulated in the accumulation means, and compares the lightness variance value for each block with a predetermined threshold value, For the block for which the brightness variance value is determined to be less than the threshold value, the number of 1 corresponding to the average brightness value of the block is read from the table, and the sign of the brightness quantized value including the number 1 equal to the number is read. It is characterized by re-creating.
[0015]
  In addition to the configuration of the image processing apparatus according to claim 1, the image processing apparatus according to claim 3
  It is characterized by further comprising color difference code requantization means for performing requantization for reducing the number of bits for the code of the average color difference value stored in the storage means for each block.
[0016]
  In addition to the configuration of the image processing device according to claim 3, the image processing device according to claim 4 includes:
  The color difference code requantization means determines whether or not the average brightness value is within a predetermined brightness range for each block, and changes a requantization method for the average color difference value according to the determination result. It is characterized by doing.
[0017]
  An image processing method according to claim 5 comprises:
  A color conversion step for converting color image data into lightness component data and color difference component data;
A lightness coding step of generating a code of an average lightness value, a lightness variance value, and a lightness quantization value by compressing the lightness component data in a predetermined block unit;
A color difference encoding step in which the color difference component data is fixed-length-compressed in units of blocks and a code of an average color difference value is generated;
An accumulating step of accumulating in the accumulating means encoded data consisting of codes of the average brightness value, the brightness variance value, the brightness quantized value, and the average color difference value;
A color-difference code removing step of removing the code of the average color-difference value stored in the storage unit in units of blocks;
A lightness code requantization step of removing the code of the lightness average value and lightness dispersion value stored in the storage means in units of blocks.
[0018]
  In addition to the configuration of the image processing method according to claim 5, the image processing method according to claim 6 includes:
  The lightness code requantization step reads out the sign of the average lightness value, lightness variance value, and lightness quantization value accumulated in the accumulation means, and compares the lightness variance value with a predetermined threshold value for each block, For a block determined to have a lightness variance value less than the threshold, the number of 1s corresponding to the average lightness value of the block from the table in which the number of 1s in the code of the lightness quantized value and the average lightness value are associated. , And the code of the lightness quantized value including the number 1 equal to the number is recreated.
[0019]
  The image processing method according to claim 7 includes, in addition to the configuration of the image processing method according to claim 5 or 6,
  A color difference code requantization step of performing requantization for reducing the number of bits for the code of the average color difference value stored in the storage means in units of blocks.
[0020]
  In addition to the configuration of the image processing method according to claim 7, the image processing method according to claim 8 includes:
  The color difference code requantization step determines whether or not an average brightness value is within a predetermined brightness range for each block, and a method of requantizing the code of the average color difference value according to the determination result It is characterized by changing.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following embodiment, color image data is divided into block units (4 × 4 pixels), and its brightness component isGThe average brightness, brightness variance, and each pixel quantized value are converted by the BTC method, and average values are generated for the color components, and the encoded data using these as codes will be described as an example, but the present invention is limited to this. Monochrome image data in the same wayGCoded data converted into average brightness, brightness variance, and pixel quantized values using the BTC method and using these as codes, and further, encoded data obtained by code-compressing color / monochrome image data using another method, generally multiple codes Needless to say, the present invention can be widely applied to fixed-length encoded data consisting of portions.
[0022]
[Encoding / Decoding]
  First, an example of color image data encoding / decoding used in the present embodiment will be described.
  FIG. 1 is a block diagram of a configuration example for compressing and encoding color image data. In the figure, 11 is a color conversion unit, 12 is a working buffer memory, 13 is a lightness encoding unit, 14 is a color difference Cr encoding unit, 15 is a color difference Cb encoding unit, and 16 is a memory device that stores encoded data. is there.
[0023]
  In the color conversion unit 11, Y (lightness), Cr, and Cb are applied to input color image data in which one pixel is R (red), G (green), and B (blue) by applying a conversion expression such as the following expression: Convert to (color difference) data.
    Y = 0.299R + 0.587G + 0.114B (1)
    Cr = 1 / (2 (1-0.299)) (R−Y) (2)
    Cb = 1 / (2 (1-0.114)) (BY) (3)
[0024]
  In the present embodiment, each of R, G, B, Y, Cr, and Cb data constituting one pixel is 8 bits (that is, quantization level 256). In the figure, for example, [7: 0] represents 8 bits of 0-7. The same applies to G, B, Y, and Cr / Cb.
[0025]
  In the buffer memory 12, Y (lightness) color-converted by the color converter 11.WhenCr,Cb (color difference) data is sequentially stored, and sequentially read out in predetermined block units (encoding units). In the present embodiment, one block is 4 × 4 = 16 pixels. Here, the position (i, j) of each 4 × 4 pixel in the block is represented by (00) to (33) (i = 0 to 3, j = 0 to 3). In the figure, for example, Y₀₀[7: 0] is Y data (8 bits) of the pixel at the position (0, 0) in one block, Y₃₃[7: 0] represents Y data (8 bits) of the pixel at the position (3, 3). Cr,The same applies to Cb data.
[0026]
  In the lightness coding unit 13, Y (lightness) data Y of 16 pixels per block is provided for each block.₀₀[7: 0] to Y₃₃From [7: 0]GApplying the BTC method, the average brightness value La [7: 0] (8 bits) of the 16 pixels in the block, the brightness dispersion value Ld[7: 0] (8 bits) and each quantization value φ of 16 pixels in the block₀₀~ Φ₃₃A code of (1 bit / pixel) is generated. Here, the average brightness value La [7: 0] is the brightness data Y of 16 pixels in the block.₀₀[7: 0] to Y₃₃An average value of [7: 0] is calculated and obtained by rounding to 8 bits. Value variance Ld[7: 0] represents the degree of lightness dispersion in the block, and Y₀₀[7: 0] to Y₃₃It is obtained by subtracting the minimum value from the maximum value in [7: 0]. Quantized value φ₀₀~ Φ₃₃Is Y with the average brightness value La as a threshold value.₀₀[7: 0] to Y₃₃Each of [7: 0] is obtained by binarizing.
[0027]
  In the chrominance Cr encoding unit 14, the chrominance data Cr of 16 pixels per block is provided for each block.₀₀[7: 0] to Cr₃₃From [7: 0], an average color difference value Cr (7: 0) (8 bits) of 16 pixels in the block is generated. Similarly, in the color difference Cr encoding unit 15, color difference data Cb of 16 pixels per block is provided for each block.₀₀[7: 0] to Cb₃₃From [7: 0], an average color difference value Cb [7: 0] (8 bits) of 16 pixels in the block is generated. Cr,The calculation of the average value of Cb is the same as in the case of La.
[0028]
  In the memory device 16, codes (φ) output from the lightness coding unit 13 and the color difference (Cr / Cb) coding units 14 and 15 in units of 4 × 4 pixel blocks.₀₀~ Φ₃₃, La [7: 0], Ld[7: 0], Cr [7: 0], Cb [7: 0]) are set as one record (48 bits), and the encoded data of the original color image data is sequentially stored.
[0029]
  2 and 3 show specific images of the GBTC encoding process in the lightness encoding unit 13. FIG. 2A shows each lightness value Y of 16 pixels per block before encoding._ij(I = 0-3, j = 0-3), and FIG. 2B shows La, Ld, φ after encoding obtained from this_ijThe value is (i = 0-3, j = 0-3). The average brightness value La = 108 of 16 pixels is equal to each brightness value Y of 16 pixels in FIG._ijIs obtained by converting the average value = 108.4375 into an integer. The lightness variance value Ld = 56 is obtained as a value obtained by subtracting the minimum value = 80 from the maximum value = 136 among the lightness values of the 16 pixels in FIG. 16 pixel quantized value φ_ijIs the brightness value Y of the corresponding pixel in FIG._ijIs obtained when La = 108 or less, and when La is greater than 108, 1 is obtained. FIG. 3 shows the distribution of the lightness values of 16 pixels in FIG.dRepresents the relationship.
[0030]
  FIG. 4 shows encoded data (φ) output in units of blocks from the lightness encoding unit 13 and the color difference (Cr / Cb) encoding units 14 and 15._ij, La, Ld, Cr, Cb). Since Y, Cr, and Cb of the original color image data are each 8 bits per pixel, in FIG. 4, as a result of encoding, a data amount of (8 + 8 + 8) × 16 = 384 bits per block of 16 pixels is obtained. It can be seen that the compression encoding is 48 bits and 1/8. Note that the encoded data of the monochrome image does not have a Cr / Cb code portion, and in the example of FIG._ij, La, Ld) only 32 bits.
[0031]
  FIG. 5 is a block diagram of a configuration example for decoding the encoded data of FIG. In the figure, 21 is a memory device, 22 is a lightness code inverse transform unit, 23 is a buffer memory, and 24 is a color inverse transform unit.
[0032]
  The memory device 21 corresponds to the memory device 16 of FIG. 1, and here, the encoded data of the bit configuration shown in FIG. 4 is stored per block. In the case of decoding, the encoded data stored in the memory device 21 is sequentially read, and the average color difference values Cr [7: 0] and Cb [7: 0] of 16 pixels in the block are stored in the buffer memory 23 in units of blocks. Although stored directly, the average brightness value La [7: 0] of the 16 pixels in the block, the brightness variance value Ld [7: 0], and the respective quantized values φ of the 16 pixels in the block₀₀~ Φ₃₃Is transmitted to the lightness code inverse transform unit 22.
[0033]
  In the lightness code inverse transform unit 22, La [7: 0], Ld [7: 0], and φ₀₀~ Φ₃₃Based on the above, each brightness value Y of 16 pixels in the block₀₀[7: 0] to Y₃₃[7: 0] is decoded. Specifically, φ_ij= 0 pixel is Y_ijIs La-Ld / 2 and φ_ij= 1 pixel is Y_ijIs La + Ld / 2.
[0034]
  In the buffer memory 23, the lightness data Y of 16 pixels in the block output from the lightness code inverse transform unit 22.₀₀[7: 0] to Y₃₃[7: 0] is stored in association with the previous average color difference data Cr [7: 0], Cb [7: 0].
[0035]
  In the color inverse conversion unit 24, each brightness data Y of 16 pixels in one block in the buffer memory 23.₀₀[7: 0] to Y₃₃[7: 0] and average color difference data Cr [7: 0] and Cb [7: 0] of 16 pixels in the block, respectively, Y [7: 0], Cr [7: 0] and Cb [ 7: 0] as a set, the following formula is applied, and R (red), G (green), B (blue) color image data R [7: 0], G [7: 0], Inversely converted to B [7: 0].
    R = Y + 2 (1−0.299) Cr (4)
    B = Y + 2 (1-0.114) Cb (5)
    G = Y + 2 (1−0.299R−0.114B) /0.587 (6)
[0036]
  Hereinafter, each example of the present invention will be described in detail. The original encoded data to be further code-compressed has the code configuration shown in FIG. La is the first code, LdIs the second code, φ_ijIs the third code, Cr,Cb is the fourth code.
[0037]
[Example 1]
  If the encoded data is intended for a normal business document image, it can be said that the code having the lowest importance among the codes is a color difference (color part) code. In the case of a photograph, if the brightness information remains, the subjectTheCan be determined. Even if it is a character / graphics part, as long as there is a difference in brightness, the meaning of the document can be retained even if the color difference information is lost. The present embodiment is based on this idea.
[0038]
  FIG. 6 is a block diagram of the first embodiment, in which 31 indicates a memory device and 32 indicates a color difference code removing unit. The memory device 31 has already stored the encoded image data. In this embodiment, encoded data having the code configuration (48 bits) shown in FIG. 4 is stored per block. The chrominance code removing unit 32 is configured to output a chrominance code portion Cr in the encoded data,A process of deleting Cb is performed.
[0039]
  FIG. 7 is a processing flowchart of the color difference code removing unit 32. In the chrominance code removing unit 32, the encoded data (La,Ld, φ_ij, Cr,Cb) is read sequentially (step 41), and the color difference code portions Cr [7: 0] and Cb [7: 0] are removed in units of one block (step 42), La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃Only the lightness code portion is rewritten in the memory device 31 (step 43). Thereafter, the processing is repeated until the processing of the processing target encoded data in the memory device 31 is completed (step 44).
[0040]
  FIG. 8 shows the code compression transition of the first embodiment. FIG. 8A shows the original encoded data of 48 bits shown in FIG. In the first embodiment, when there is a request for further code compression of the color image data compressed and encoded as shown in FIG. 8A, the color difference code portion Cr,Cb is removed. As a result, as shown in FIG. 8B, there are 32 bits per block, which is even smaller than in FIG. Since the GBTC inverse conversion is not performed again on the compression code of (a) and the conversion is not performed again, the code length can be easily shortened. If each code data is consolidated for each block (for example, as a record) and stored at consecutive addresses in the memory device, the code of the next block is sequentially shifted by the code length of the deleted color difference to obtain consecutive addresses. Can be freed up. Even if the code not to be deleted is not shifted, the address to be deleted may be released so that other data can be stored.
[0041]
  Decoding processing from the compression-encoded data in FIG. 8B may be performed with Cr = Cb = 0 in the previous equations (4), (5), and (6).
[0042]
[Example 2]
  In this embodiment, GBTC encoded data (La, Ld, φ) of only the lightness portion from which the color difference code portion is deleted in the first embodiment._ij) From La,Ld is removed and the lightness quantization value φ_ijThis is the only sign. Lightness quantization value φ_ijOutputs the re-quantized (re-encoded) lightness distribution Ld. This requantized lightness quantization value is expressed as φ ′_ijI will call it. Note that GBTC encoded data of monochrome image data is originally La, Ld, φ_ijTherefore, when code compression is further performed, the second embodiment is directly applied.
[0043]
  FIG. 9 is a block diagram of the second embodiment, in which 51 is a memory device, 52 is a lightness code requantization unit, and 53 is a φ ′ creation table. The memory device 51 has already accumulated image data that has been subjected to code compression. In this embodiment, GBTC encoded data of only the lightness code portion of the code configuration shown in FIG. 8B is accumulated per block. The lightness code requantization unit 52 calculates Lac from the GBTC encoded data of only the lightness code part.,While deleting Ld, the lightness quantization value φ ′_ij(I = 0-3, j = 0-3).
[0044]
  FIG. 10 is a processing flowchart of the lightness code requantization unit 52. In the lightness code requantization unit 52, only GBTC encoded data (La, Ld, φ) stored in the memory device 51 is stored._ij) Is read (step 61), and its brightness variance Ld[7: 0] is compared with a predetermined threshold T (for example, T = 128) (step 62). If Ld ≧ T, the current lightness quantization value φ₀₀~ Φ₃₃As it is φ ′₀₀~ Φ '₃₃And La,Ld is removed (step 63), φ₀₀~ Φ₃₃= Φ ′₀₀~ Φ '₃₃Are rewritten to the memory device 51 (step 65). On the other hand, Ld <TIn this case, referring to the φ ′ creation table 53, φ ′₀₀~ Φ '₃₃As well as La,Ld is removed (step 64) and φ ′₀₀~ Φ '₃₃Are rewritten to the memory device 51 (step 65). Thereafter, the processing is repeated until all the processing of the encoded data to be processed in the memory device 51 is completed (step 66).
[0045]
  FIG. 11 shows the code compression transition of the second embodiment. FIG. 11A shows 32-bit GBTC encoded data compressed by applying the first embodiment as shown in FIG. 8B. When this is further compressed by applying the second embodiment, as shown in FIG. 11 (b), it becomes 16 bits per block, which is half that of FIG. 11 (a). Moreover, it becomes 1/3 from the original 48-bit encoded data in FIG. In the monochrome image data, the original GBTC-encoded data has the code configuration of FIG. 11A, and when the second embodiment is applied, similarly to FIG. 11B, it becomes half.
[0046]
  The decoding process from the encoded data in FIG._ij= 0 pixel is Y_ij= 0, φ '_ij= 1 pixel Y_ij= 255, Cr = Cb = 0. In this case, the decoded image is a complete binary image.
[0047]
  FIG. 12 shows an example of specific processing in steps 63 and 64 of FIG. 10, and FIG. 13 shows an example of the φ ′ creation table 53 used for the processing.
[0048]
  A block having a large lightness variance value Ld (Ld is equal to or greater than the threshold value T) has a large edge. This is, for example, an area where characters are written, so even if the image is binarized, the meaning of the image document is rarely lost. FIG. 12A shows an example. In this case, φ_ij= Φ ′_ijAnd
[0049]
  Meanwhile, figure12 (b) and (c) show the case where Ld is less than the threshold value T. Such a block is a region where there is no edge and is gentle, for example, a region where a photograph is placed. Therefore, even in a binary image, it can be said that image quality degradation is small if the gradation, that is, the average brightness value La is expressed by the number of black pixels in the block. Therefore, first, the number of black pixels in the block is determined by the table of FIG. 13 for each value of La. This black pixel count is φ_ijIf the number of pixels is equal to 1, φ_ijΦ ′_ijAnd If they are different, φ_ij= 1 forcibly match the number of black pixels. In the case of FIG. 12B, since La = 108, the number of black pixels in the block is set to 6 from the table of FIG. However, φ_ij= 1 has 9 pixels, so φ for 3 pixels_ij= 1 is changed to 0. Here, follow the raster order from the upper left pixel of the block,_ijChange φ ′_ijHave created. On the other hand, FIG._ijThis is an example of changing = 0 to 1.
[0050]
Example 3
  FIG. 14 is a block diagram of the third embodiment. In the figure, 70 is a memory device, 72 is a code compression device, and the code compression device 72 comprises a color difference code removal unit 73, a lightness code requantization unit 74, and a φ ′ creation table 75. That is, this is a combination of Example 1 and Example 2. FIG. 15 shows the transition of code compression in this embodiment.
[0051]
  First, the memory device 70 stores color image data compressed into encoded data of 48 bits per block (4 × 4 pixels) as shown in FIG. When it is necessary to further compress this image data, first, the color difference code removing unit 73 removes the 32-bit color difference code portion per block as shown in FIG. The code is compressed into a code (the code in FIG. 8B) and rewritten in the memory device 70. If this is insufficient and further compression is necessary, the lightness code requantization unit 74 then converts the GBTC code of FIG. 15B stored in the memory device 70 into the further FIG. ) Quantization value φ ′ having 16 bits per block as shown in FIG._ijThe code is compressed to the only code and rewritten in the memory device 70.
[0052]
  According to the third embodiment, the free capacity of the memory device that stores the encoded image data can be improved stepwise and adaptively. For example, when various kinds of various image data are stored in the memory device, the compression processing is made up to FIG. 15 (b) or continued to FIG. 15 (c) depending on the type of image data or the like. It can be used properly, for example, as in (a).
[0053]
  In monochrome document image data, the original encoded data is a GBTC code having only a lightness portion as shown in FIG. 15B. Therefore, when it is necessary to further compress the encoded data, The lightness code requantization unit 74 may be activated.
[0054]
Example 4
  The first to third embodiments are basically based on the assumption that the color difference code portion in the encoded data is removed. However, in the fourth embodiment, when further compression of the encoded data is required, First, this color difference code portion is reduced and replaced.
[0055]
  FIG. 16 is a block diagram of the fourth embodiment, in which 81 is a memory device, and 82 is a color difference code requantization unit. Memory device81Corresponds to the memory device 16 of FIG. 1 or the memory device 21 of FIG. 5 and stores image data of 48 bits of encoded data per block (4 × 4 pixels) as shown in FIG. Here, Cr and Cb are each 8 bits, and the quantization level is 255 (-127 to +127). The color difference code requantization unit 82 performs a process of reducing the number of bits by requantizing (recoding) the color difference code portions Cr and Cb in the encoded data. Here, it is assumed that the quantization is requantized to 7 levels, and the 16-bit color difference code portion of Cr [7: 0] and Cb [7: 0] is reduced to a total of 6 bits of C [5: 0].
[0056]
  FIG. 17 is a process flowchart of the color difference code requantization unit 82. The chrominance code requantization unit 82 sequentially reads out the encoded data stored in the memory device 81 (step 91), and the codes La [7: 0], Ld [7: 0], φ in units of one block.₀₀~ Φ₃₃, Cr [7: 0] and Cb [7: 0], the color difference code portions Cr [7: 0] and Cb [7: 0] are requantized, and both are combined into C [5: 0]. (Step 92), encoded data La [7: 0], Ld [7: 0], φ again₀₀~ Φ₃₃, C [5: 0], the data is rewritten in the memory device 81 (step 93). Thereafter, the processing is repeated until all the processing of the encoded data to be processed in the memory device 81 is completed (step 94). Here, the re-quantization process in step 92 can be most simply realized by dividing the values of Cr and Cb by a predetermined number (for example, 32) and rounding them to 3 bits. Further, it may be indexed by a requantization table. In this case, since the output value C is directly obtained corresponding to the input values of Cr and Cb, division and rounding can be omitted.
[0057]
  FIG. 18 shows the transition of code compression when this embodiment is applied. FIG. 18A shows the original encoded data of 48 bits per block shown in FIG. When this embodiment 4 is applied, this encoded data becomes 38 bits per block as shown in FIG. The code in FIG. 18B is not subjected to a reverse GBTC conversion to the code compression in FIG. 18A, and is not converted again, so that the code length can be easily shortened. If each code is fixed for each block and stored at consecutive addresses in the memory device, the code of the next block is shifted sequentially by the code length of the reduced color difference to make the consecutive address free space. Can do. Even if the code not to be deleted is not shifted, the address to be deleted may be released so that other data can be stored.
[0058]
  The decoding process from the encoded data in FIG. 18B is performed by using the first 3 bits of C [5: 0] as Cr and the latter 3 bits as Cb in equations (4), (5), and (6). Just do it.
[0059]
  FIG. 19 shows an image of color difference code requantization processing according to this embodiment. In the figure, thin vertical and horizontal lines represent threshold values for requantization, and by converting the values that have entered each region surrounded by the quantization threshold values into respective quantization representative values indicated by black circles, Cr, It shows that Cb is re-quantized and re-encoded to reduce the code. Here, in FIG. 19A, the quantization levels 255 (-127 to +127) of Cr and Cb are requantized and re-encoded to 7 levels (for example, each of Cr and Cb is divided by 32 and rounded). ), A code of 16 bits in total with Cr and Cb is reduced to a total of 6 bits.
[0060]
  The requantization method is not limited to this. For example, since the color difference space of Cr and Cb is a visually non-uniform space, it is possible to visually reduce more effectively by setting the quantization threshold line non-uniformly. This can be easily realized by using a table. FIG. 19 shows an example of scalar quantization for Cr and Cb. For example, two-dimensional vector quantization in which adjacent regions are integrated and one quantization representative value is assigned is more effective. It becomes reduction.
[0061]
  By the way, actual Cr and Cb may not exist in the entire range of FIG. A region surrounded by an ellipse in FIG. 19 schematically represents the range. Therefore, it is more effective that the quantized representative value is placed only within this range. However, this range moves with the brightness value. Wide in medium brightness and narrow in shadow and highlight areas. Therefore, La is used as the brightness value, and the quantization threshold value and the quantization representative value are moved according to the value of La. FIG. 19B shows that the quantization is performed in a narrower range depending on the difference in brightness value. FIG. 19B shows an example in which Cr and Cb are each divided by 16.
[0062]
  In FIG. 20, the color difference requantization method as shown in FIG. 19A is applied to the intermediate brightness, and the color difference requantization method as shown in FIG. 19B is applied to the shadow / highlight portion. The process flowchart of the color difference code requantization part 82 in the case is shown. The chrominance code requantization unit 82 reads the encoded data from the memory device 81 (step 101), and 1 block of codes La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃, Cr [7: 0] and Cb [7: 0] are determined, it is determined whether the value of the average brightness code La [7: 0] is within the threshold values T1 and T2 (step 102). In the case of T1 ≦ La <T2 (intermediate lightness), the color difference codes Cr [7: 0] and Cb [7: 0] are requantized and recoded by applying FIG. 19A (Cr and Cb are 32). And round them together) to obtain C [5: 0] (step 103). On the other hand, when La <T1 (shadow) or La> T2 (highlight), Cr [7: 0] and Cb [7: 0] are requantized and recoded by applying FIG. 19B. (Cr, Cb is divided by 16 and rounded), and C5: 0] (step 104). The encoded data La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃, C [5: 0] are rewritten in the memory device 81 (step 105). Thereafter, the processing is repeated until all the processing target data in the memory device 81 is processed (step 106).
[0063]
Example 5
  This is a combination of Example 4 and Example 1. That is, according to the fourth embodiment, first, the original encoded data La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃, Cr [7: 0] and Cb [7: 0], the color difference code portions Cr [7: 0] and Cb [7: 0] are reduced and replaced with C [5: 0], and the encoded data La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃, C [5: 0]. If this is still insufficient, then C [5: 0] that has been reduced and replaced is completely removed, and the encoded data is converted into La [7: 0], Ld [7: 0], φ₀₀~ Φ₃₃This is a code for only the lightness part of. The configuration and the code transition state of the fifth embodiment will be described together in the sixth embodiment, and will be omitted here.
[0064]
Example 6
  FIG. 21 is a block diagram of the sixth embodiment. In the figure, 110 is a memory device, 120 is a code compression device, and the code compression device 120 includes a color difference code requantization unit 121, a color difference code removal unit 122, a lightness code requantization unit 123, and a φ ′ creation table 124. Become. That is, the sixth embodiment is configured by combining the fifth embodiment and the second embodiment. In the drawing, the configurations of the color difference code requantization unit 121 and the color difference code removal unit 122 correspond to the fifth embodiment. The code compression transition in the sixth embodiment is shown in FIG.
[0065]
  In the memory device 110, first, image data compressed into 48-bit encoded data per block (4 × 4 pixels) as shown in FIG. When it is necessary to further compress the image data in the memory device 110, first, the color difference code requantization unit 121 performs color difference code Cr.,Cb total 16 bits are re-quantized (re-encoded) to color difference code C total 6 bits, and compressed to 38 bits of encoded data per block as shown in FIG. Rewrite. When it is necessary to further compress the document image data in the code format of FIG. 22B rewritten and stored in the memory device 110, the color difference code removing unit 122 next performs the color difference code C6 bit. 22 is completely removed, and as shown in FIG. 22C, the data is compressed into a GBTC code having only a lightness code portion of 32 bits per block and rewritten in the memory device 110. Then, when it is necessary to further compress the document image data in the code format of FIG. 22 (c) rewritten and accumulated in the memory device 110, the brightness code requantization unit 123 then This time La of the lightness code part,Remove Ld total 16 bits and φ_ijIs re-quantized (re-encoded), and as shown in FIG. 22 (d), the lightness re-quantized value φ ′ of 16 bits per block_ijThe code is compressed to the only code (binarized code) and rewritten in the memory device 110.
[0066]
  According to the present embodiment, the free capacity of the memory device for storing the image data can be improved stepwise and adaptively as compared with the third embodiment.
[0067]
[Application example]
  Here, as an example, an example of application to an image processing apparatus such as a filing system that handles a large amount of document image data is shown. Of course, the code compression method of the present invention can be applied to various other image processing apparatuses and systems. It is.
[0068]
  FIG. 23 is a block diagram showing an embodiment of an image processing apparatus such as a document image filing system to which the code compression method of the present invention is applied. The image processing apparatus includes a display device 200, a keyboard 210, a mouse (pointing device) 220, a scanner 230, a printer 240, a document file 250, a processing control device (CPU) 260, an encoding / decoding processing device 270, a code compression device 280, The document image database 290, the secondary storage device 300, and a bus 310 for connecting them are included.
[0069]
  The CPU 260 controls each unit and registers / searches the document image database 280. The encoding / decoding processing device 270 is configured as described with reference to FIGS. The code compression apparatus 280 may have any of the configurations described in the first to sixth embodiments. Here, the configuration of the best mode of the sixth embodiment (see FIG.21). As will be described later, the CPU 260 controls which code compression unit of the code compression device 280 is operated. The code compression device 280 may be detachable from the CPU 260, for example, as a board (board) structure, or may be built in from the beginning together with the code / decoding processing device 270.
[0070]
  The image data of the original document is read by the image scanner 230 and transferred to the CPU 260 as image data. The existing image data of the floppy disk, MO, and other document files 250 are directly transferred to the CPU 260. Although omitted in FIG. 1, the image data may be received from a remote terminal through a communication line, and similarly transferred to the CPU 260. The CPU 260 applies preprocessing such as editing to the input image data if necessary and passes it to the encoding / decoding processing device 270. The encoding / decoding processing device 270 encodes the image data and stores it in the document image database 290. In this way, the database 290 accumulates a large amount of encoded data of document image data for each document.
[0071]
  The user inputs a data search request for the document image database 290 using the keyboard 210 or the like. When there is a data search request, the CPU 260 reads the encoded data of the corresponding document from the document image database 280, passes it to the encoding / decoding processing device 270, decodes it by the encoding / decoding processing device 270, and displays it on the display device 200. . When there is a print request, the printer 240 prints out.
[0072]
  The above is the general operation of this type of image processing apparatus. Hereinafter, characteristic operations of the image processing apparatus equipped with the code compression apparatus 280 will be described.
  Here, in the document image database 290, the color difference quantization value φ obtained by encoding the document image data in one block unit (4 × 4 pixels), and the brightness portion is encoded by the GBTC method as shown in FIGS. 4 and 22A._ij16 bits, average brightness value La8 bits, brightness variance value Ld8-bit, color difference average value Cr for color components,It is assumed that encoded data of 48 bits in total of 8 bits for each Cb is accumulated.
[0073]
  The CPU 260 manages the amount of data stored in the document image database 290, and displays a document list of image data stored in the document image database 290 on the display device 200 when the allowable capacity is exceeded. The user selects one or a plurality of document image data that may be further compressed from the document list display using the keyboard 210 or the mouse 220, and requests code compression. Here, when the document image data is a normal business document image, the code having the lowest importance among the codes is the code of the color difference, and can retain the meaning of the document even if the color difference information is somewhat lost. .
[0074]
  Upon receiving the code compression request, the CPU 260 first activates the color difference code requantization unit 281 in the code compression device 280 to instruct code compression of the selected document image data. The color difference requantization unit 281 reads the instructed document image data from the database 290, code-compresses the encoded data as shown in FIG. 22A as shown in FIG. 22B, and rewrites it into the database 290. . As a result, the document image data is reduced from 48 bits to 38 bits per block, and a margin is generated in the database 290 accordingly. In the document image database 290, the document image data is managed as color difference requantized data.
[0075]
  If the accumulated data amount in the document image database 290 still exceeds the allowable capacity even after the code compression process, or if the accumulated data amount again exceeds the allowable capacity after registration for a while, the CPU 260 again Is inquired about document image data to be code-compressed. As a result, when the user selects new document image data, the CPU 260 activates the color difference code requantization unit 281 in the code compression device 280 again. On the other hand, when selecting the document image data that has undergone color difference requantization, the CPU 260 then activates the color difference code removal unit 282 to instruct the code compression of the document image data that has undergone color difference requantization. The color difference code removing unit 282 reads the instructed document image data from the database 290, further code-compresses the document image data that has been code-compressed as shown in FIG. 22B, and the database 290 as shown in FIG. Rewrite to. As a result, the document image data is reduced from 38 bits to 32 bits per block. The document image database 290 manages the document image data as color difference code deleted data.
[0076]
  Thereafter, when the same operation is repeated and the color difference code deleted data is selected, the CPU 260 next activates the lightness code requantization unit 283 in the code compression device 280, and the color difference code deleted image data is deleted. Instructs code compression. The lightness code requantization unit 283 reads the instructed document image data from the database 290, and further code-compresses the code-compressed document image data as shown in FIG. 22C, as shown in FIG. Rewrite to database 280. As a result, the document image data is reduced to 16 bits per block. The document image database 290 manages the document image data as color difference code deleted / lightness requantized data.
[0077]
  FIG. 24 shows a control flowchart of the CPU 260 for the above operation. The monochrome document image data is originally stored in the database 290 in the code format shown in FIG. Therefore, when the monochrome document data is selected as the code compression target, the CPU 260 may activate the lightness requantization unit 283 in the code compression device 280 from the beginning.
[0078]
  In the above operation, whenever the usage state of the document image database 290 exceeds the allowable capacity, the CPU 260 inquires the user, and based on the code compression state of the selected document image data, the color difference code requantization in the code compression device 280 is performed. The CPU 260 activates any one of the conversion unit 281, the color difference code removal unit 282, and the lightness code requantization unit 283, but the CPU 260 selects the selected document image until the use data amount of the document image database 290 becomes less than the allowable capacity. For data, the color difference code requantization unit 281, the color difference code removal unit 282, and the lightness code requantization unit 283 in the code compression device 280 can be automatically started in order. FIG. 25 shows a control flowchart of the CPU 260 in this case. Whether the interactive format as shown in FIG. 24 or the automatic activation as shown in FIG. 25 is used may be instructed by the user in advance at the time of the first selection, for example.
[0079]
  24 and 25, the document image database 290 generally includes the document image data originally encoded by the encoding / decoding processing device 270, the code compression units 281, 282, and 283 of the code compression device 280. Further, code-compressed color difference code requantized document image data, color difference code deleted document image data, and color difference code deleted / lightness code requantized document image data are mixed.
[0080]
  At the time of retrieval, in the decoding process of the encoding / decoding processing device 270, the original encoded document image data may be decoded as shown in FIG. On the other hand, the color difference code requantized document image data is decoded with the first half of the requantized color difference code C as Cr and the second half as Cb in equations (4), (5), and (6). In this case, the color information is degenerated but not lost at all. The document image data from which the color difference code has been deleted is decoded as Cr = 0 and Cb = 0. This is the same as the code of the monochrome document image data, and the color information is lost, but the meaning of the document can be retained in a normal business document image. The color difference code deleted / lightness code requantized document image data has a lightness requantization value φ ′_ij= 0 pixel is Y_ij= 0 (white), φ '_ij= 1 pixel Y_ij= 256 (black), Cr = 0, Cb = 0. In this case, the decoded image is a complete binary image. Depending on the document, the meaning of the document can still be retained.
[0081]
  Next, use of the secondary storage device 300 of FIG. 23 will be described. When the user selects document image data to be code-compressed and requests the CPU 260 to perform code compression processing, if the user wants to leave the original encoded data of the document image, the user instructs the CPU 260 to that effect. When the CPU 260 receives this instruction, the encoded document image data is read from the document image database 290 and stored in the secondary storage device 300 prior to the activation of the code compression device 280. As a result, when the decoded image from the code-compressed data in the document image database 290 is insufficient at the time of search, the original encoded data in the secondary storage device 300 is read and decoded by the encoding / decoding processing device 270. Thus, an original image can be obtained.
[0082]
〔recoding media〕
  The code compression method of the present invention can be provided by being recorded on a computer-readable recording medium such as a floppy disk, a CD-ROM, or a memory card as a computer-executable program. The image processing apparatus as shown in FIG. 23 is constructed using the hardware and software resources of a so-called computer system. In this case, the code compression apparatus 280 is installed by installing the program of the recording medium. Has a predetermined function. Note that the program for the encoding / decoding processing device 270 may be recorded together on the recording medium.
[0083]
【The invention's effect】
  According to the image compression method, the image compression apparatus, and the image processing apparatus of the present invention, the following effects can be obtained.
(1) Since only a part of information that is not important for image quality is deleted or reduced and replaced, it is possible to easily realize a process for improving the compression rate for encoded data.
(2) Since the code length per block is always constant even after the compression rate improvement processing, the ease of processing / editing with respect to the compressed code is not lost.
(3) By discarding information that is not important for image quality in order, a high-quality image can be restored at each stage. Since multistage processing is adaptive processing, high image quality according to various image characteristics can be obtained.
(4) Since the code length per block in each stage is always constant, as described in (2), the compression code can be easily processed and edited, and the compression rate multi-stage processing itself is also simple. It has become a thing.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration example of an encoding mechanism that generates encoded data used in an embodiment of the present invention.
FIG. 2 is a diagram illustrating a specific example of encoding processing.
3 is a diagram showing a distribution of each pixel value in FIG. 2. FIG.
FIG. 4 is a diagram showing a bit configuration of encoded data used in an embodiment of the present invention.
FIG. 5 is a block diagram illustrating a configuration example of a decoding mechanism that decodes encoded data used in an embodiment of the present invention.
FIG. 6 is a block diagram of a first embodiment of code compression according to the present invention.
FIG. 7 is a process flowchart of the first embodiment of the present invention.
FIG. 8 is a diagram illustrating a transition of code compression according to the first embodiment of this invention.
FIG. 9 is a block diagram of a second embodiment of code compression according to the present invention.
FIG. 10 is a process flowchart of the second embodiment of the present invention.
FIG. 11 is a diagram illustrating a transition of code compression according to the second exemplary embodiment of the present invention.
FIG. 12 is a diagram showing a specific processing example of the second exemplary embodiment of the present invention.
FIG. 13 is a diagram showing an example of a lightness requantized value creation table used in the second embodiment of the present invention.
FIG. 14 is a block diagram of a third embodiment of code compression according to the present invention.
FIG. 15 is a diagram illustrating a transition of code compression according to the third exemplary embodiment of the present invention.
FIG. 16 is a block diagram of a fourth embodiment of code compression according to the present invention.
FIG. 17 is a process flowchart of the fourth embodiment of the present invention.
FIG. 18 is a diagram illustrating a transition of code compression according to the fourth exemplary embodiment of the present invention.
FIG. 19 is a conceptual diagram of code compression according to a fourth embodiment of the present invention.
FIG. 20 is another processing flowchart of the fourth embodiment of the present invention.
FIG. 21 is a block diagram of a fifth embodiment of code compression according to the present invention.
FIG. 22 is a diagram illustrating a transition of code compression according to the fifth exemplary embodiment of the present invention.
FIG. 23 is an overall block diagram of a document image filing system according to an application example of the present invention.
FIG. 24 is a flowchart for explaining an operation example of the document image filing system of FIG.
FIG. 25 is a flowchart for explaining another example of the operation of the document image filing system of FIG.
[Explanation of symbols]
  200 Display device
  210 keyboard
  220 mice
  230 Scanner
  240 Printer
  250 document files
  260 Processing control unit (CPU)
  270 encoding / decoding processing device
  280 Code compression device
  281 Color difference code requantization unit
  282 Color difference code removal unit
  283 Lightness Code Requantizer
  290 Document Image Database
  300 Secondary storage

Claims (8)

Color conversion means for converting color image data into lightness component data and color difference component data;
  Lightness encoding means for compressing the lightness component data in a predetermined block unit and generating codes of average lightness value, lightness variance value, and lightness quantization value;
  Color difference encoding means for generating a code of an average color difference value by compressing the color difference component data at a fixed length in block units;
  Accumulation means for accumulating encoded data composed of codes of the average lightness value, the lightness variance value, the lightness quantization value, and the average color difference value;
  Color difference code removing means for removing the code of the average color difference value stored in the storage means in units of blocks;
  Lightness code requantization means for removing the code of the lightness average value and lightness variance value stored in the storage means for each block;
An image processing apparatus comprising: The lightness code requantization means includes a table in which the number of 1s in the code of the lightness quantization value and the average lightness value are associated with each other.
  The lightness code requantization means reads the sign of the average lightness value, lightness variance value, and lightness quantization value accumulated in the accumulation means, and compares the lightness variance value for each block with a predetermined threshold value, For the block for which the brightness variance value is determined to be less than the threshold, the number of 1 corresponding to the average brightness value of the block is read from the table, and the sign of the brightness quantized value including the number 1 equal to the number is read. The image processing apparatus according to claim 1, wherein the image processing apparatus is recreated. The color difference code requantization means for performing requantization for reducing the number of bits for the code of the average color difference value accumulated in the accumulation means for each block. 2. The image processing apparatus according to 2. The color difference code requantization means determines whether or not an average brightness value is within a predetermined brightness range for each block, and requantizes a code of the average color difference value according to the determination result The image processing apparatus according to claim 3, wherein: A color conversion step for converting color image data into lightness component data and color difference component data;
  A lightness coding step of generating a code of an average lightness value, a lightness variance value, and a lightness quantization value by compressing the lightness component data in a predetermined block unit;
  A color difference encoding step for generating a code of an average color difference value by compressing the color difference component data in a fixed length in units of blocks; and
  An accumulating step of accumulating in the accumulating means encoded data consisting of codes of the average brightness value, the brightness variance value, the brightness quantized value, and the average color difference value;
  A color-difference code removing step of removing the code of the average color-difference value stored in the storage unit in units of blocks;
  A lightness code requantization step of removing the code of the lightness average value and lightness dispersion value stored in the storage means in units of blocks;
An image processing method comprising: The lightness code requantization step reads out the sign of the average lightness value, lightness variance value, and lightness quantization value accumulated in the accumulation means, and compares the lightness variance value with a predetermined threshold value for each block, For a block determined to have a lightness variance value less than the threshold, the number of 1s corresponding to the average lightness value of the block from a table in which the number of 1s in the code of the lightness quantization value and the average lightness value are associated The image processing method according to claim 5, further comprising: re-creating a code of a lightness quantization value including a number of 1 equal to the number. For the sign of the average color difference value stored in the storage means, 7. The image processing method according to claim 5, further comprising a color difference code requantization step of performing requantization for reducing the number of dots in units of blocks. The color difference code requantization step determines whether or not an average brightness value is within a predetermined brightness range for each block, and a method of requantizing the code of the average color difference value according to the determination result The image processing method according to claim 7, wherein:

Images