JP2002232721A - Image coding apparatus and its method, and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce a coding amount that are generated when coding an image that includes regions coded by different compression rates to below a predetermined amount, and to perform the coding considering the image quality. SOLUTION: A coding amount B (1) that are generated in a time coding is obtained (S1001). Then, assuming a=-2/3 an approximate expression representing a relationship between the coding amount and a quantization parameter is determined (S1002). A target value is then substituted in the approximate expression to obtain a quantization parameter at that time for use in the second time coding (S1003). When the coding amount does not reach the target value even at the second time coding (S1000), a coding amount B (2) at that time and B (1) are substituted in the approximate expression to determine an approximate expression (S1005). The target value is substituted in the approximate expression to recalculate the value of the quantization parameter (S1006) for another execution of quantization and coding using the recalculated quantization parameter.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention encodes an image including regions to be encoded at different compression ratios,
The present invention relates to an image encoding device and a method for controlling a code amount, and a storage medium.

[0002]

2. Description of the Related Art Conventional color still image compression methods include:
The JPEG method using discrete cosine transform and the method using Wavelet transform are used. This type of encoding method basically performs uniform processing on the entire image. Therefore, when the encoding method is changed for each image partial area, image area flag information for identifying the area is required separately. Become.

The purpose of dividing image areas in image compression is to perform the following processing.

(1) The character area and the photographic area are divided into areas, and the quantization step of the photographic area where image quality degradation is relatively inconspicuous is coarsened.

[0005] The quantization of a character area where image quality deterioration is relatively conspicuous is made fine.

(2) The natural image part and the computer graphics part are divided into regions, and the quantization step of the natural image region where image quality deterioration is relatively inconspicuous is coarsened.

[0007] The quantization of a computer graphics area where image quality degradation is relatively conspicuous is made finer.

By switching the encoding process in this way, the image quality can be improved with a smaller code amount.

In the conventional technique, when the generated code amount exceeds a predetermined code amount, the original is read again and the DC is read.
A method of changing the quantization parameter of the T coefficient and performing recompression has been employed.

[0010]

However, the above-described processing performed when the generated code amount exceeds a predetermined code amount is for the entire image area, and is estimated from the entire code amount. Was divided into regions (for example, a high compression region and a low compression region), it was not possible to cope with encoding changed for each region. In particular, no consideration was given to the balance between the code amount and the image quality of the two regions as described above.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and it is an object of the present invention to reduce the amount of generated code to a predetermined amount or less when coding images including regions to be coded at different compression rates. With the goal. It is another object of the present invention to perform encoding in consideration of image quality during the encoding.

[0012]

In order to achieve the object of the present invention, for example, an image coding apparatus of the present invention has the following arrangement.

That is, an image coding apparatus for coding an image including regions to be coded at different compression rates, comprising: a calculating means for obtaining a code amount for each region; a code amount and a quantization parameter; Estimating means for estimating a quantization parameter corresponding to the code amount by the calculation means for each of the regions, so that the code amount calculated by the calculation means is equal to or less than a predetermined amount. The estimating means estimates a quantization parameter for each of the regions.

Further, there is provided a dividing means for dividing the image into tiles having a predetermined size.

[0015] Further, encoding is performed on the original image using a plurality of different quantization parameters, and is further decoded to create information relating to the distortion of the obtained image with respect to the original image and the code amount of the image. Means for estimating a quantization parameter according to distortion allowed for each of the regions using the information.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail according to preferred embodiments with reference to the accompanying drawings.

[First Embodiment] FIG. 2 is a block diagram showing the configuration of a digital image device as an image encoding device according to this embodiment to which an image scanner and a page description language rendering unit are connected. In the figure, reference numeral 120 denotes an image scanner, and 1001 denotes an image scanner 120.
The image data 1003 is input from the image scanner 120 to the selector 99, and the image area flag 1003 is input from the image scanner 120 to the selector 99.

Reference numeral 121 denotes a page description language rendering unit, 1002 denotes image data input from the page description language rendering unit 121 to the selector 99, and 1004 denotes an image area flag described later input from the page description language rendering unit 121 to the selector 99. It is.

Reference numeral 99 denotes a selector, 100 denotes an image input unit, 1
01 is a line buffer for compression processing, 102 is an image area flag encoding unit, 103 is an (image) encoding unit, 125 is a monitoring unit, 104 is a compression buffer, 105 is a hard disk (HDD), 106 is a decoding buffer, 107 is an image area flag decoding unit, 108 is an (image) decoding unit, and 109 is a line buffer for decoding processing. Reference numeral 110 denotes an interface (I / F) with a printer or the like, which is an image generation unit (not shown).

The components of the digital imaging device according to the present embodiment having the above-described configuration will be described below.

The image input unit 100 receives image data and an image area flag signal from the image scanner 120 selected by the selector or the page description language rendering unit 121. The image data input to the image input unit 100 is divided into tiles in the line buffer 101. In this embodiment, since image data is divided into tiles, it is assumed that an image area flag is provided for each tile. Hereinafter, the image data for each tile is referred to as tile image data.

When the image data is divided into tiles (hereinafter, the size of the tile is assumed to be M), a discrete cosine transform coding (JPEG), which is coding of color information for each M × M pixel, is performed by a coding unit. At 103, run-length encoding is performed on the image area flag by the image area flag encoding unit 102. At that time, the encoding process of the tile image data in the encoding unit 103 is switched in accordance with the input of the image area flag (uncoded signal) from the image area flag encoding unit 102. That is,
Switch encoding parameters. Details will be described later. Further, the monitoring unit 125 constantly monitors the generated code amount, and when the generated code amount exceeds a preset amount,
An instruction is given to the image scanner 120 to reread the manuscript or the like, or the page description language rendering unit 12
1 is instructed to re-output the rendering result.

FIG. 3 is a block diagram showing the internal configuration of the encoding unit 103 and the decoding unit 108, and showing encoding and decoding processing. In the figure, reference numeral 11 denotes a color conversion unit which converts RGB signals included in an input image into a luminance / color difference signal (YCbC
r). In the present embodiment, the RGB signals are color signals and have 256 gradations, but are not limited thereto.

Reference numeral 12 denotes a discrete cosine transform (DCT) unit.
A spatial frequency transform (DCT transform) is performed on each of the luminance / color difference signals, and DCT coefficients corresponding to the respective signals are output. Reference numeral 13 denotes a quantization unit that reduces a data amount by quantizing DCT coefficients. 14 is a variable length encoder (VL
In the section C), the quantized DCT coefficients (quantized values) are subjected to Huffman coding to further reduce the data amount. The above is each part in the encoding part 103.

A memory 15 stores compressed image data and compressed image area flag data. 16 is a variable length decoding unit (VL
In D), Huffman decoding is performed on the compressed image data. Reference numeral 17 denotes an inverse quantization unit that returns the DCT coefficient value. 18 is I
The DCT unit performs a DCT inverse transform to return to a luminance / color difference signal. Reference numeral 19 denotes a color conversion unit which returns the luminance / color difference signals to RGB signals. The above-described variable length decoding unit 16, inverse quantization unit 17, IDC
The T unit 18 and the color conversion unit 19 are each unit of the decoding unit 108.

The monitoring unit 125 monitors the generated code amount in the encoding unit 103. If the generated code amount is expected to exceed a preset value, the image scanner 120 or the page description language rendering is performed as described above. The above-described instruction is output to the unit 121 and the quantization unit 13
Is instructed to switch the quantization step.

On the other hand, the image area flag encoding unit 102 encodes the image area flag as described above and records it in the memory 15. The image area flag decoding unit 107 decodes the encoded image area flag, and switches a quantization matrix for inverse quantization of image data according to the decoding result.
Output to

FIG. 1 is a flowchart of an encoding process for each block image data in this embodiment. In addition,
Prior to the following processing, a target average code amount (details will be described later) and an initial value of a quantization parameter for each area (high compression rate area and low compression rate area) are set. Then, the following processing is performed on all block image data. When a high-compression region and a low-compression-ratio region are mixed in one tile, a region having a larger ratio in the tile is set as the attribute of the tile. For example, if a high compression rate area and a low compression rate area are mixed in a tile and the ratio of the high compression rate area to this tile is larger than that of the low compression rate area, this tile is included in the high compression rate area. And

First, the attribute of this tile image is determined from the image area flag data for each block image data (step S11). As a result, if the tile image data is a photographic area (or a natural image area), the process proceeds to step S12, and encoding at a high compression rate described later is performed (step S11). On the other hand, if the tile image data is a character area (or a computer graphics area), the process shifts to step S13 to perform encoding at a low compression rate, which will be described later (step S1).
3). Next, the code amount per pixel (average code amount) is calculated from the number of pixels in the tile (step S14). If the average code amount is larger than a predetermined average code amount (step S15), the process proceeds to step S16. Calculate the quantization parameter of the block image data. The processing according to the above flowchart is performed on all block image data.

Here, it is known that there are the following approximate expressions for the code amount and the quantization parameter.

Log10 (B) = a × log10 (F) + b (1) where B is the average code amount per pixel (Bitrate)
(Bits / pixel), F is a quantization parameter, and its default value is 50. a and b are predetermined constants that determine the property of (1).

FIG. 7 is an explanatory diagram of the relationship between the quantization parameter F and the average code amount B. In this figure, the equation (1) of the code amount B and the quantization parameter F is fixed at a = − ／ and b is changed to 1, 1.5 and 2.
As shown in the figure, the larger the F, the smaller the average code amount B tends to be, and the larger the b, the larger the average code amount B tends to be.

From the above, when two or more average code amounts B for the quantization parameter F are measured (of course, two or more quantization parameters F are used), a and b are determined from the approximate expression (1). Since the approximate expression (1) is determined, the quantization parameter F to be set at that time can be estimated from the target average code amount B. Hereinafter, the process will be described with reference to the flowchart of the process of estimating the quantization parameter F of the present embodiment shown in FIG.

In this embodiment, the code amount B (1) generated in the first encoding processing is first obtained (step S100).
1). Then, b is obtained assuming that a = −ｂ, and the approximate expression (1) is determined (step S1002). Then, the target value B0 is put into the equation (1), and the value of the quantization parameter F at that time is obtained, and used for the second encoding process (step S1003). The target value will be described later.

If the code amount does not reach the target value even in the second encoding process (step S1000), the above 2
The code amount B (2) generated in the second encoding is calculated (step S1004), and the two sample points B (1) and B (2) are substituted into the approximate expression (1) to obtain coefficient values a and b. Is obtained (step S1005). Here, since the approximate expression (1) is determined, the target value B0 is substituted into the approximate expression (1), and the value of the quantization parameter F at this time is recalculated (Step S).
1006) The quantization and the encoding are performed again using the recalculated quantization parameter F.

The processing according to the flowchart shown in FIG. 10 is the content of the processing in step S16 in the processing shown in FIG.

In this case, the code amount becomes the expected amount in most cases. However, if the value of the quantization parameter F obtained by the second calculation for obtaining the quantization parameter F is changed to a slightly larger value, 3 As a result of the second recalculation of the quantization parameter F, the code amount at the time of encoding with the obtained quantization parameter F becomes equal to or less than the predetermined code amount with a high probability. That is, in the flowchart shown in FIG.
The variable n is an integer of at most 3 or less.

By executing the above processing, each area (high compression rate area and low compression rate area) can be compressed at a target value or less.

In the present embodiment, the change value of the quantization parameter for each area of the image (high compression area and low compression area) is converted into the code amount for each tile generated in the past scan. We further propose a method of estimating based on
In particular, a method of determining two areas (high compression area and low compression area) in consideration of the balance between the code amount and the image quality will be described.

First, the area ratio of the character region (the ratio of the number of tiles including the character region to the total number of tiles) is α, and the area ratio of the photograph region (the ratio of the number of tiles including the photograph region to the total number of tiles) is ( 1−α), and the average code amount of the character portion is B
Assuming that 1 (the average value of the average code amount of the tiles as the text portion) and the average code amount of the photo portion is B2 (the average value of the average code amounts of the tiles as the photo portion), the target bit rate (target value) Bt can be obtained by the following equation.

B1 · α + B2 · (1−α) = Bt (2) When a character area and a photographic area are mixed in one tile, an area having a larger ratio in the tile is defined as the tile. Attribute. For example, when a character area and a photograph area are mixed in a tile and the ratio of the photograph area to the tile is larger than that of the character area, the tile is assumed to be included in the photograph area.

In this embodiment, before performing the third recalculation of the quantization parameter F, a decoded image obtained by encoding and decoding the original image data using a plurality of quantization parameters is used. An average square error is calculated for each pixel value of the original image data and the decoded image data, and the result is obtained as a rate distortion curve as shown in FIG.

In the rate distortion curve shown in FIG. 8, the allowable distortion D1 of the character area and the allowable distortion D2 of the photographic area are determined from the target bit rate B (horizontal axis) and the allowable distortion D (vertical axis). Target bit rate B1, target bit rate B2
Is determined with reference to the rate distortion curve shown in FIG. If the bit rate becomes equal to or less than the target bit rate in the equation (2), the quantization parameter F of each area can be determined from FIG.

However, in the case of exceeding the target, by balancing D1 and D2, B1,
A process for determining B2 is performed.

(1) B1 is obtained from the lower limit of D1, which is a character area in which deterioration is conspicuous, and is calculated backward from Bt to obtain B1.
2. Find D2.

(2) B2 is obtained from the lower limit of D2, which is a photographic area where deterioration is not conspicuous, and is calculated backward from Bt to obtain B2.
1, D1 is obtained.

The magnitude of the distortion of D2 of (1) and the magnitude of D1 of (2) are compared to determine which method to use.
The decision formula is determined experimentally.

FIG. 11 shows a flowchart of the above processing, which will be described below. The processing according to this flowchart is the flowchart shown in FIG.
0, executed when n is 3 or more. (So,
The processing in steps S1004, S1005, and S1006 is executed only when n = 2. First, a rate distortion curve is created as described above (step S110).
1). Then, allowable distortions D1 and D2 for each area (high compression rate area and low compression rate area) are set (step S1102). The settings of D1 and D2 are not limited to being set in this step, but may be set in advance.

Then, B1 and B2 for D1 and D2 are calculated using the rate distortion curve set in step S1101.
Is obtained (step S1103). Then, with reference to the target value Bt, B1 and B2 are set (adjusted) so as to be equal to or less than the target value (step S1104). And B1, B
A quantization parameter corresponding to 2 is obtained from equation (1) (step S1105).

FIG. 4 is a diagram showing an example of a character area obtained as a result of performing area division on an original image. In the case of PDL rendering input, the area corresponds to a character area and a photograph area. . Also, in the case of a scanner input, it is shown that the region is divided into a region having a relatively small quantization step and a region having a relatively large quantization step according to the region determination result such as a character region or a photograph region. I have. The quantization parameter F is switched for each tile included in each area, and each code amount is counted.

FIG. 5 is a diagram showing an encoding method applied to each tile in the original image divided into tiles. In the figure, a region where the quantization parameter is small is shown in white, and a region where the quantization parameter is large is shown in gray.

FIG. 6 is an example of a basic quantization matrix for 8 × 8 DCT coefficients. When the quantization parameter F is obtained, the basic matrix value is multiplied by F / 50 to form an actual quantization step, and the DCT coefficient is quantized (divided) in this quantization step.

On the other hand, in the inverse quantization, this quantization step is multiplied by the Huffman decoded value to return to the DCT coefficient value.

As described above, with the image encoding apparatus and method according to the present embodiment, the code amount is controlled to be equal to or less than the target value even for an image divided into a high compression rate area and a low compression rate area. Becomes possible. At this time, encoding can be performed in consideration of image quality.

[Second Embodiment] In the present embodiment, a rate distortion curve determiner shown in FIG. 9 is newly provided in the apparatus shown in FIG. As a result, the number of rereads can be reduced.

Reference numeral 301 denotes a discrete cosine transform (DCT)
Device, 302 is an inverse quantizer, 303 is an inverse DCT device, 304
Is a mean square error calculator, 305 is a variable length code amount calculator, 306 is a generator, and 309 is an image data signal.

Here, 302a, 303a, 304a,
Reference numeral 305a denotes data obtained by omitting the lower two bits of the DCT coefficient and performing linear quantization 1/4 (F = 200). 302b, 303b, 304b, 305
b is data obtained by omitting the lower one bit of the DCT coefficient and subjected to linear quantization (1/2) (F = 100).

Note that 302a, 303a, 304a, 3
05a, 302b, 303b, 3
The data input to 04b and 305b may be opposite to each other.

The two code amounts are calculated. Further, a mean square error of the difference between the decoded image and the original image, which has been subjected to inverse quantization and inverse DCT, is calculated. Thereby, the code amount and the mean square error (MSE) are obtained at two points, and the curve corresponding to FIG. 8 can be approximated. This is effective in reducing the code amount and the arithmetic processing of the mean square error (MSE). As a result, the rate distortion curve generator 306 generates the curve of FIG. 8, and D for input B and B for input D are obtained.

[0060]

As described above, according to the present invention, when coding an image including regions to be coded at different compression ratios, the amount of generated codes can be reduced to a predetermined amount or less. In this encoding, encoding can be performed in consideration of image quality.

[Brief description of the drawings]

FIG. 1 is a flowchart of an encoding process for each block image data according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a digital image device as an image encoding device according to the first embodiment of the present invention, to which an image scanner and a page description language rendering unit are connected.

FIG. 3 is a block diagram showing an internal configuration of an encoding unit 103 and a decoding unit 108 and showing encoding and decoding processing.

FIG. 4 is a diagram illustrating an example of a character area obtained as a result of performing area division on an original image.

FIG. 5 is a diagram illustrating an encoding method performed on each tile in an original image obtained by dividing a tile.

FIG. 6 is a diagram illustrating an example of a basic quantization matrix for an 8 × 8 DCT coefficient;

FIG. 7 is an explanatory diagram of a relationship between a quantization parameter F and an average code amount B.

FIG. 8 is a diagram showing a rate distortion curve.

FIG. 9 is a diagram illustrating a configuration of a rate distortion curve determiner according to a second embodiment of the present invention.

FIG. 10 is a flowchart of a process of estimating a quantization parameter F according to the first embodiment of the present invention.

FIG. 11 is a flowchart illustrating a process of estimating a quantization parameter F according to the first embodiment of the present invention.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shinichi Kato 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (72) Inventor Kenichi Ota 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon In-house F term (reference) 5C059 KK25 MA23 MC14 ME02 ME05 PP14 PP19 SS12 SS20 TA47 TB08 TC02 TC04 TC06 TC18 TC24 TC38 TD03 TD05 TD06 TD12 UA02 UA39 5C078 AA04 BA44 BA57 DA01 DA02 DB07 5J064 AA02 BA09 BC13 BC

Claims (19)

[Claims] 1. An image coding apparatus for coding an image including regions to be coded at different compression rates, comprising: a calculating unit for obtaining a code amount for each region; Estimating means for estimating, for each of the regions, a quantization parameter corresponding to the code amount by the calculating means, using the relational expression of: The image coding apparatus according to claim 1, wherein the estimating unit estimates a quantization parameter for each of the regions. 2. The image encoding apparatus according to claim 1, further comprising a division unit configured to divide the image into tiles having a predetermined size. 3. The image coding apparatus according to claim 2, wherein said calculating means calculates a code amount for each area for each tile, and said estimating means estimates a quantization parameter for each tile. . 4. An original image is coded using a plurality of different quantization parameters, and further decoded, and information on distortion of the obtained image with respect to the original image and the code amount of the image is created. 4. The apparatus according to claim 1, further comprising: a creating unit configured to estimate a quantization parameter corresponding to distortion allowed for each region using the information. 5. Image coding device. 5. The image according to claim 4, wherein the creation unit creates the information when the code amount calculated by the calculation unit does not fall below the predetermined amount twice or more. Encoding device. 6. The method according to claim 1, wherein the estimating unit specifies a code amount corresponding to distortion allowed for each area using the information, and estimates a quantization parameter corresponding to the code amount using the relational expression. The image encoding device according to claim 4 or 5, wherein 7. The information processing method according to claim 4, wherein the information is information on a curve indicating a relationship between the code amount and a distortion of the image having the code amount with respect to the original image. The image encoding device according to claim 1. 8. The apparatus according to claim 1, wherein the area is classified into an area encoded at a high compression rate and an area encoded at a low compression rate. The image encoding device according to claim 1. 9. The image encoding apparatus according to claim 8, wherein the area encoded at a high compression rate includes a photograph area. 10. The image encoding apparatus according to claim 8, wherein the area encoded at a low compression rate includes a character area. 11. The relational expression is log10 (B) = a
× log10 (F) + b, where B is the average code amount, F
The image encoding apparatus according to claim 1, wherein is a quantization parameter, and a and b are predetermined constants. 12. The apparatus according to claim 1, wherein the code amount is an average code amount per pixel.
Item 7. The image encoding device according to Item 1. 13. The image encoding apparatus according to claim 1, wherein the encoding uses a discrete cosine transform. 14. An image coding method for coding an image including regions to be coded at different compression rates, comprising: a calculating step of calculating a code amount for each region; And an estimation step of estimating a quantization parameter corresponding to the code amount in the calculation step for each of the regions using the relational expression of the above, so that the code amount calculated in the calculation step is equal to or less than a predetermined amount. The image encoding method according to claim 1, wherein the estimating step estimates a quantization parameter for each of the regions. 15. The image encoding method according to claim 14, further comprising a dividing step of dividing the image into tiles having a predetermined size. 16. An original image is coded using a plurality of different quantization parameters, and further decoded, and information on distortion of the obtained image with respect to the original image and the code amount of the image is created. The image encoding method according to claim 14, further comprising: estimating a quantization parameter corresponding to distortion allowed for each region using the information in the estimation step. 17. A computer-readable storage medium storing a program code for executing an image encoding method for encoding an image including an area to be encoded at a different compression rate, wherein the computer-readable storage medium stores And a program code for an estimation step for estimating, for each region, a quantization parameter corresponding to the code amount in the calculation step, using a relational expression between the code amount and the quantization parameter. A computer-readable storage medium, comprising: estimating a quantization parameter for each of the regions in the estimation step such that a code amount calculated in the calculation step is equal to or less than a predetermined amount. 18. The computer-readable storage medium according to claim 14, further comprising a program code of a dividing step of dividing the image into tiles having a predetermined size. 19. An original image is coded using a plurality of different quantization parameters, and further decoded, and information on distortion of the obtained image with respect to the original image and the code amount of the image is created. The computer according to claim 17 or 18, further comprising: a program code of a creating step of: estimating a quantization parameter corresponding to distortion allowed for each region using the information in the estimation step. A readable storage medium.