JP2003032680A - Image coded, image coding method, program code of the same image coder, and memory medium of the same image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the difference between the image quality of each tile and the one of input image of a image coder, even when decoding the coded data of each tile. SOLUTION: A code-quantity control portion 104 of the image code so decides every tile divided by a tile dividing portion 101 the basic code-quantity to be assigned to each tile as to control every tile the increase/decrease of the basic code-quantity to be assigned to each tile, by using the information of the flatness calculated every tile by a tile-flatness calculating portion 17.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus for coding an image, an image coding method, a program code, and a storage medium.

[0002]

2. Description of the Related Art JPEG2000 image coding system (IS
O / IEC FDIS15444-1), a coding method for Wavelet transforming and coding an image is higher than a JPEG baseline coding method using block-based DCT even if the degree of compression is increased. There is no block distortion and there is a merit in image quality. In order to make the most of such image quality advantage, it is effective to perform Wavelet conversion on the entire input image. However, when actually implementing the encoding device, the size of the input image is extremely large. Due to a large size or a limited buffer capacity of the encoding device, the entire input image cannot be wavelet-converted and encoded at one time, and the input image is divided into a plurality of regions (tiles), The technique of encoding each tile may be taken.

An example of such tile division processing is shown in FIG. In the figure, the input image 21 is divided into tiles 22 and tiles 23 by a tile dividing part 24 at a dotted line portion, and the divided tiles 22 and 23 are encoded by an encoder one by one. Since the divided tiles 22 and 23 are smaller in size than the input image 21, the buffer capacity of the subsequent encoder is limited, which allows easier and low-cost implementation.

FIG. 3 shows a functional configuration of an image coding apparatus that divides an input image into tiles and encodes each divided tile.

A tile dividing unit 301 divides the input image into regions (tiles) having a predetermined size, and outputs the divided tiles to the transform coding unit 302 one by one. 302
Is a transform coding unit, which transforms the input tiles, generates transform coefficients, and outputs them to the quantization / entropy coding unit 303. This conversion processing includes, for example, frequency conversion, and discrete wavelet conversion or D
There are CT etc. Reference numeral 303 denotes the quantization / entropy coding unit 3
At 03, first, the transform coefficient is quantized to obtain the quantized value of the transform coefficient. Then, the quantized value is entropy-encoded and output as encoded data. A header is attached to the encoded data, and the header includes parameters used in the above-described encoding process, such as a quantization step and a tile size. Three
Reference numeral 04 is a code amount control unit, which is a quantization / entropy encoding unit 3
In 03, the code amount generated when performing the quantization and entropy coding is monitored, and the quantization / entropy coding unit 303 is controlled according to the generated code amount (coding control).
To do. For example, the size of the generated encoded data is set to a size designated in advance.

As a method of controlling the code amount, ISO / I
EC FDIS15444-1 "Rate cont
The "roll" section uses a rate / distortion optimization method that minimizes the error between the encoded image and the input image at the specified code amount, and the best image quality is obtained in the sense that the error is the minimum. The generated code amount is controlled so that

[0007]

When the code amount control is performed by using the rate / distortion optimization method, the control is such that the best image quality is obtained in the sense that the error is minimum. However, due to human visual characteristics, the minimum error does not always correspond to the visually good image quality. In particular, in the case of performing tile division and encoding, the following problems occur only by minimizing the error. This problem will be described using the input image 21 shown in FIG.

First, the input image 21 is compared with the tile 22 containing many flat image areas and the tile 23 containing many non-flat image areas. Here, the flat image area is
An area including pixels in which the change in each pixel value is relatively small is shown. On the other hand, an image area that is not flat indicates an area that includes pixels in which the change in each pixel value is relatively large.

ISO / IEC FDIS15444-1
When the code amount control described in the above is performed, the code amount control is performed so as to minimize the error with respect to the entire image.
2, the code amount is optimally distributed to the tile 23 in terms of error. However, when actually looking at the image after decoding, the image quality of the tile 23 that is not flat is relatively good, whereas the image quality of the tile 22 that is flat becomes blurry compared to the tile 23, and the entire image is viewed. Then, an image with a sense of discomfort is obtained in which there is a difference in image quality between the flat portion and the non-flat portion. This is due to the effect of the masking effect on human visual characteristics.In a non-flat image area, a coding error is masked by fluctuations in the image information itself, and even a slightly large error is inconspicuous, whereas in a flat area, masking works. This is because even a small error is noticeable because it is not present.

The present invention has been made in view of the above problems, and it is an object of the present invention to reduce the difference in image quality between tiles even when the encoded data of each tile is decoded.

[0011]

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

That is, an image coding apparatus for coding an image in accordance with the JPEG2000 (ISO / IEC 15444-1) standard, which comprises a dividing means for dividing an image into tiles, and a conversion process for each tile Transform means for generating a coefficient, encoding means for performing an encoding process on the transform coefficient, basic code amount calculation means for calculating a basic code amount to be assigned to each tile, and attention based on the data of the tile of interest An index calculation unit for calculating an index indicating the flatness of the tile, and an index indicating the flatness of the target tile calculated by the index calculation unit for the basic code amount to be assigned to the target tile calculated by the basic code amount calculation unit. Accordingly, the correction code amount calculation means for increasing or decreasing the allocation amount such that the allocation amount increases as the flatness increases, and the allocation of the tile of interest calculated by the correction code amount calculation means. And a control means for controlling the encoding means on the basis of the issue amount.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will now be described in detail according to preferred embodiments with reference to the accompanying drawings.

[First Embodiment] FIG. 1 shows the functional arrangement of an image coding apparatus according to this embodiment.

A tile division unit 101 divides an input image into regions (tiles) of a predetermined size, and outputs each tile to a transform coding unit 102.

A transform coding unit 102 transforms each tile to generate transform coefficients. In this embodiment, frequency conversion is used as this conversion processing. The frequency conversion includes DCT and discrete wavelet conversion, but any of these may be used.

Reference numeral 103 denotes a quantization / entropy coding unit,
The transform coefficient by the transform coding unit 102 is quantized using a predetermined quantization step, and entropy coding is performed on the quantized transform coefficient.

A code amount control unit 104 monitors the code amount of the coded data generated by the quantization / entropy coding unit and the entropy coding unit, and quantizes / quantizes based on this code amount. The entropy coding unit 104 is controlled. Details will be described later.

Reference numeral 105 denotes a tile flatness calculation unit, which calculates the flatness for each tile and uses the calculated flatness as the code amount control unit 10.
Output to 4. Details will be described later.

FIG. 5 shows the basic configuration of the image coding apparatus according to this embodiment.

Reference numeral 501 is a CPU, which is a RAM 502 or ROM.
The program and data stored in 503 are used to control the entire image encoding apparatus, and also execute the image encoding processing of this embodiment having the functional configuration shown in FIG.

A RAM 502 has an area for temporarily storing programs and data loaded from the external storage device 504 and the storage medium drive 509, and has a C
The PU 501 has a work area used when executing various processes.

A ROM 503 stores programs and data for controlling the entire image coding apparatus.

Reference numeral 504 denotes an external storage device such as a hard disk, which can store programs and data loaded from the storage medium drive 509. Further, when the size of the work area exceeds the capacity of the RAM 502, the excess area can be provided as a file.

Reference numerals 505 and 506 are a keyboard and a mouse.
Various instructions can be input to the image encoding device.

A display device 507 is composed of a CRT, a liquid crystal screen, etc., and can display various image information and character information.

An image input device 508 is composed of a scanner, a digital camera and the like, and can input an image as digital data to the image coding device. The image input device includes an interface for connecting to the image encoding device.

A storage medium drive 509 is a CD-RO.
Programs and data can be read from a storage medium such as an M, a DVD-ROM, and a floppy (registered trademark) disk, and output to the RAM 502 and the external storage device 504.

A bus 510 connects the above-mentioned units.

The code amount control unit 104 and the tile flatness calculation unit 1 of the image coding apparatus of the present embodiment having the above configuration.
The processing in 05 will be described in detail below.

The code amount control unit 104 is, for example, an ISO / IEC
A basic code amount to be assigned to each tile is determined by a method such as rate / distortion optimization described in FDIS15444-1. Then, using the flatness information for each tile calculated by the tile flatness calculation unit 17, control is performed to increase or decrease the basic code amount to be allocated for each tile.

The flatness of tile i is given by Fi (F
Assuming that the larger i is, the flatter the tile is), the basic code amount allocation B _base-i for each tile is increased or decreased according to Fi as shown in the following equation, and the code amount Bi actually allocated to the tile i is Can be obtained.

Bi = B _base−i × (1 + f (Fi)) Here, the function f (x) is a function that monotonically increases with respect to the value of x. For example, f (x) = a (x− b) However, a and b are constants, and a> 0 is considered. In the case of the function of this example, the value of the constant a determines the degree of increase or decrease of the code allocation amount Bi with respect to the change amount of the flatness Fi, and if the value of the flatness Fi is larger than the value of the constant b, f (Fi) > 0, that is, Bi is increased and Fi is increased.
When the value of is smaller than the value of b, f (Fi) <0, that is, Bi is decreased. The values of the constants a and b are determined so that a visual improvement effect can be obtained and the code amount of the entire image becomes the intended code amount.

The function f (x) is not limited to the function of this example as long as it is a function that increases monotonically, but for a very large value of x or a very small value of x, the code allocation amount Bi In order to avoid an extreme change in the value, it is desirable to saturate the value of the function f (x) at the upper limit or the lower limit for such x.

As the flatness Fi calculated by the tile flatness calculation unit 105, the reciprocal of the variance of each pixel value in the tile i is used, but the invention is not limited to this, and the pixel value in each tile It is also possible to obtain the sum of absolute values of the differences between the average values of the pixel values and use the reciprocal thereof. Alternatively, when each pixel value in the tile is sequentially scanned, the sum of the absolute values of the differences between two adjacent pixel values may be obtained and the reciprocal thereof may be used as the flatness.

The pixel value in tile i is set to P _ij (0 ≦ j <N
-1, N is the number of pixels in tile i), and _P'is the average value of P _ij , and when the inverse of the variance of each pixel value in tile i is used as the flatness Fi, the flatness is calculated according to the following equation. F
i can be obtained.

Fi = N / {(P _i0 -P ') ² + (P
_i1− P ′) ² ,, + (P _{i (N−1)} −P ′) ² } On the other hand, as the flatness Fi, the absolute value sum of the difference between each pixel value in the tile and the average value of the pixel values is calculated. , And the reciprocal thereof, the flatness Fi can be obtained according to the following equation.

Fi = N / {| P _i0 −P ′ | + | P _i1
−P ′ | ,, + | P _{i (N−1)} −P ′ |} Or, as the flatness Fi, when scanning each pixel value in the tile in order, the absolute difference between two adjacent pixel values is obtained. When the sum of the values is calculated and its reciprocal is used, the flatness Fi can be calculated according to the following equation.

Fi = N / {| P _i1 −P _i0 | + | P
_{i2- Pi1} | ,, + | _{Pi (N-1 )} -P
_{i (N−2)} |} Then, using the flatness Fi obtained by the tile flatness calculation unit 105 as described above, the code amount control unit 104 sets the code amount Bi actually assigned to each tile into the above formula. Seek based on.

When all tiles are coded by the above processing, coded data including the coding result of each tile is generated. A header including a quantization step for each tile and parameters used for entropy coding is attached to this coded data.

As described above, the image coding apparatus and the image coding method according to the present embodiment perform coding in consideration of the flatness of each tile. Therefore, even if these tiles are decoded, each tile is decoded. The difference in image quality can be reduced.

[Second Embodiment] A case where wavelet transform is used as the frequency transform process in the transform coding unit 101 will be described. When the transform coding unit 101 performs wavelet transform for each tile, each tile is divided into subbands as shown in FIG.

In the figure, since the coefficient of the LL subband corresponds to the reduced image of the tile, the flatness of the LL subband can be calculated without calculating the flatness of the entire tile. If done, the flatness of the tile can be calculated with a small amount of calculation. Specifically, when the coefficient value of the LL subband of tile i is LL _ik (0 ≦ k <M−1, M is the number of transform coefficients in the LL subband), and the average value of LL _ik is LL ′, When the reciprocal of the variance of each transform coefficient in the tile i is used as the flatness Fi, the flatness Fi can be obtained according to the following formula.

Fi = M / {(LL _i0 -LL ') ² +
(LL _i1 −LL ′) ² , + (LL _{i (M−1)} −L
L ′) ² } On the other hand, as the flatness Fi, the absolute value of the difference between each transform coefficient in the subband LL and the average value of the transform coefficients is obtained, and when the reciprocal thereof is used, the flatness Fi is obtained according to the following formula. be able to.

Fi = M / {| LL _i0 −LL ′ | + | L
L _i1 −LL ′ |, + | LL _{i (N −1)} −LL ′ |} Alternatively, when the respective transform coefficient values in the subband LL are sequentially scanned as the flatness Fi, two adjacent transforms are obtained. When the sum of the absolute values of the difference between coefficient values is calculated and the reciprocal thereof is used, the flatness Fi can be calculated according to the following formula.

Fi = N / {| LL_i1-LL_i0│ + │
LL_i2-LL_i1| ,, + | LL _{i (N-1)}-LL
_{i (N-2)}|}

As described above, according to the present invention, even when the encoded data of each tile is decoded, the difference in image quality between tiles can be reduced.

[Brief description of drawings]

FIG. 1 is a diagram showing a functional configuration of an image encoding device according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating a processing example of tile division.

FIG. 3 is a diagram showing a functional configuration of a conventional image encoding device that divides an input image into tiles and encodes each divided tile.

FIG. 4 is a diagram showing subbands.

FIG. 5 is a diagram showing a basic configuration of an image encoding device according to an embodiment of the present invention.

Claims (20)

[Claims] 1. An image coding apparatus for coding an image in accordance with the JPEG2000 (ISO / IEC 15444-1) standard, wherein the image coding apparatus divides the image into tiles, and a conversion process is performed on each tile. Transform means for generating a coefficient, encoding means for performing an encoding process on the transform coefficient, basic code amount calculation means for calculating a basic code amount to be assigned to each tile, and attention based on the data of the tile of interest An index calculation unit for calculating an index indicating the flatness of the tile, and an index indicating the flatness of the target tile calculated by the index calculation unit for the basic code amount to be assigned to the target tile calculated by the basic code amount calculation unit. Accordingly, the correction code amount calculating means for increasing or decreasing the allocation amount so that the allocation amount increases as the flatness increases, and the allocation code amount of the target tile calculated by the correction code amount calculating means. The image encoding device characterized by comprising a control means for controlling the encoding means are. 2. The image coding apparatus according to claim 1, wherein the index calculation means uses an inverse number of a variance of the data of the tile of interest as an index of the tile of interest. 3. The index calculating means obtains an absolute sum of differences between each data of the tile of interest and an average value of the data, and uses the reciprocal thereof as an index of the tile of interest. 1. The image encoding device according to 1. 4. The index calculating means obtains a total sum of absolute values of differences between two adjacent data when scanning each data of the tile of interest, and uses the reciprocal thereof as an index of the tile of interest. The image coding apparatus according to claim 1, wherein: 5. The image coding apparatus according to claim 1, wherein the data of the tile of interest is a value of a pixel included in the tile of interest. 6. The image coding apparatus according to claim 1, wherein the data of the tile of interest is a transform coefficient forming an LL subband of the tile of interest. 7. The basic code amount calculation means calculates a basic code amount to be assigned to the tile of interest as a rate / rate.
The image coding apparatus according to claim 1, wherein the image coding apparatus is obtained by using distortion optimization. 8. The index of the target tile is F, the basic code amount to be allocated to the target tile is B _base ,
When the function f (x) is a function that monotonically increases with respect to the value of x, the modified code amount calculation means assigns the code amount B assigned to the target tile to B = B _base × (1 + f (F) )
The image coding apparatus according to claim 1, wherein the image coding apparatus obtains by calculating. 9. The modified code amount means is characterized in that f (x) saturates to an upper limit value for a large value of x and f (x) saturates to a lower limit value for a small value of x. The image encoding device according to claim 8. 10. An image encoding method for encoding an image according to the JPEG2000 (ISO / IEC 15444-1) standard, comprising a dividing step of dividing an image into tiles, and a conversion process for each tile A conversion process for generating coefficients, an encoding process for performing the encoding process on the conversion coefficient, a basic code amount calculation process for calculating a basic code amount to be assigned to each tile, An index calculation step of calculating an index indicating the flatness of the tile, and a basic code amount to be assigned to the target tile calculated in the basic code amount calculation step, an index indicating the flatness of the target tile calculated in the index calculation step. Accordingly, the correction code amount calculation step of increasing or decreasing the allocation amount so that the allocation amount increases as the flatness increases, and the allocation code amount of the target tile calculated in the correction code amount calculation step Picture coding method and a controlling step of controlling No. step. 11. The image coding method according to claim 10, wherein in the index calculating step, an inverse number of a variance of the data of the tile of interest is used as an index of the tile of interest. 12. The index calculating step obtains an absolute sum of differences between each data of the target tile and an average value of the respective data, and uses the reciprocal thereof as an index of the target tile. 10. The image coding method according to item 10. 13. In the index calculation step, when each data of the tile of interest is scanned in order, a sum of absolute values of differences between two adjacent data is calculated, and the reciprocal thereof is used as an index of the tile of interest. 11. The method according to claim 10,
The image coding method described in. 14. The image encoding method according to claim 10, wherein the data of the tile of interest is a value of a pixel included in the tile of interest. 15. The image encoding method according to claim 10, wherein the data of the tile of interest is a transform coefficient forming an LL subband of the tile of interest. 16. In the code amount calculation step, a basic code amount to be assigned to the tile of interest is set as a rate / rate.
The image coding method according to claim 10, wherein the image coding is performed using distortion optimization. 17. When the index of the tile of interest is F, the basic code amount assigned to the tile of interest is B _base , and the function f (x) is a function that monotonically increases with respect to the value of x, In the correction code amount calculation step, the code amount B to be assigned to the target tile is B = B _base × (1 + f
The image encoding method according to claim 10, wherein the image encoding method is obtained by calculating (F)). 18. The modified code amount means is characterized in that f (x) saturates to an upper limit value for a large value of x and f (x) saturates to a lower limit value for a small value of x. The image coding method according to claim 17. 19. A program code for executing the image coding method according to claim 10. Description: 20. A computer-readable storage medium storing the program code according to claim 19.