JP2013153257A - Image processing device and image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device and method that select an optimum image compression algorithm for compression processing with high accuracy from among a plurality of image compression algorithms by a simple calculation.SOLUTION: An image processing device 1 includes: an image division section 11 for dividing an image into macro blocks; an algorithm selection section 12 for recognizing an image state of each macro block and selecting an optimum image compression algorithm for each macro block on the basis of the image state; and an image compression section 13 for compressing each macro block by the image compression algorithm selected by the algorithm selection section 12. The algorithm selection section 12 includes a plurality of image compression algorithms, and selects the optimum image compression algorithm by using as an index of evaluation the slope of a tangent to an arbitrary point on a graph parameterized by image quality and compressibility under the compression of each macro block by each image compression algorithm.

Description

  The present invention relates to an image processing apparatus and an image processing method for selecting an image compression algorithm optimal for compression processing from a plurality of image compression algorithms.

  2. Description of the Related Art Conventionally, there is a technique for dividing an image into units composed of a set of a plurality of pixels called macroblocks and performing various image processing on image data in units of macroblocks. A similar technique is used in image compression processing. Conventionally, there is a method for selecting an optimal image processing algorithm for each macroblock based on the result of squaring the quantization coefficient in this case (for example, see Patent Document 1).

Special table 2010-521113

  However, a method based on the result of squaring the quantized coefficient often complicates the calculation procedure, and can provide a good selection accuracy in selecting an optimal algorithm for compression processing. There is a problem that it is often difficult.

  The present invention has been made in view of such problems, and an image processing apparatus and an image processing method capable of selecting an image compression algorithm optimal for compression processing with high accuracy from among a plurality of image compression algorithms by simple calculation. It is an issue to provide.

  In order to solve this problem, the invention described in claim 1 includes an image dividing unit that divides an image into macroblocks, an algorithm selecting unit that selects an optimal image compression algorithm for each macroblock, and the algorithm selection. An image compression unit that performs compression processing for each macroblock using the compression algorithm selected by the unit, wherein the algorithm selection unit includes a plurality of image compression algorithms, and the macro For each block, the optimum image compression algorithm is selected by using, as an evaluation index, the slope of the tangent of an arbitrary point in the graph using the image quality and compression rate when compressed by the image compression algorithm as parameters. It is characterized by that.

ただし、
Ｓ：評価の指標
λ：任意の係数
Ｑ：現在注目しているマクロブロックにおける、圧縮前の画像データに対する、圧縮後展開時に再現された画像データにおいて再現された前記画像データの２乗誤差
Ｃ：現在注目しているマクロブロックにおける、圧縮時の符号量
ｃ_１：画質と圧縮率とをパラメータとするグラフにおける、任意の画像圧縮アルゴリズムによって形成される相関曲線の任意の点における接線の傾き
Blocks：評価のために順次選択されたマクロブロックのブロック数

According to a second aspect of the present invention, in addition to the configuration according to the first aspect, the algorithm selecting unit compresses the image quality after compression by the image compression algorithm for each macroblock and the compressed image. The code amount is used in the following formulas (1) and (2), and the one with the smallest or largest evaluation index value in the formula (1) is selected as the optimum image compression algorithm. And
S = λQ + C (1)

However,
S: Evaluation index λ: Arbitrary coefficient Q: Square error C of the image data reproduced in the image data reproduced at the time of decompression with respect to the image data before compression in the macro block currently focused on C: Code amount c ₁ at the time of compression in the macroblock currently focused on: slope of tangent at an arbitrary point of a correlation curve formed by an arbitrary image compression algorithm in a graph using image quality and compression rate as parameters
Blocks: Number of macroblocks sequentially selected for evaluation

According to a third aspect of the present invention, in addition to the configuration according to the second aspect, the algorithm selection unit is configured to sequentially store the value of the c ₁ in the equation (2) recorded in the algorithm selection unit in advance. The image compression algorithm is selected using at least _one of the values of c1 in the equation (2) calculated by calculation.

According to a fourth aspect of the present invention, in addition to the configuration according to the second aspect, the algorithm selection means uses the c ₁ in the equation (2) calculated by sequential calculation, and the processed macroblock Using the value of c ₁ generated by processing with an arbitrary quality parameter.

  The invention according to claim 5 is an image division procedure for dividing an image into macroblocks;

  An image comprising an algorithm selection procedure for selecting an optimal image compression algorithm for each macroblock, and an image compression procedure for performing compression processing for each macroblock using the image compression algorithm selected in the algorithm selection procedure In the algorithm selection procedure, for each macroblock, the slope of a tangent line at an arbitrary point in a graph using the image quality and the compression ratio as parameters for each macroblock is evaluated. By using this as an index, the optimum image compression algorithm is selected from a plurality of image compression algorithms.

  According to the first and fifth aspects of the present invention, for each macroblock, the slope of the tangent of an arbitrary point in the graph using the image quality and the compression ratio as parameters for each macroblock is compressed. By selecting the optimal image compression algorithm by using it as an evaluation index, the optimal image compression algorithm for each macroblock is uniquely determined by calculation, always taking into account the state of image quality and code amount. Can do. Also, simple mathematical formulas are sufficient. In addition, by performing such selection sequentially for all macroblocks, an optimal image compression algorithm can be applied to each macroblock.

  As described above, according to the first and fifth aspects of the invention, it is possible to select an image compression algorithm optimum for compression processing from a plurality of image compression algorithms with high accuracy by simple calculation. .

  According to the second aspect of the present invention, the numerical values obtained based on the image quality and the code amount when the plurality of image compression algorithms are respectively applied to the currently focused macroblock are respectively compared, and the smallest numerical value or the largest numerical value is compared. Is selected as the optimal image compression algorithm, the numerical values obtained by actually applying each image compression algorithm to the macroblock are compared, and the optimal image compression algorithm is selected based on the simple selection of numerical values. You can choose. This makes it possible to select an optimal image compression algorithm for the compression process simply and reliably based on specific calculations.

According to the third aspect of the present invention, by selecting an image compression algorithm using c ₁ in Equation (2) recorded in advance in the algorithm selection unit, the value of the slope c ₁ can be quickly obtained with a small amount of calculation. Can be requested. Further, by selecting an image compression algorithm using the value of c ₁ in the calculated formulas (2) by the sequential calculation, it is possible to obtain a high value of the slope c ₁ accuracy.

According to the fourth aspect of the present invention, the value of the inclination c ₁ can be obtained with higher accuracy.

1 is a functional block diagram illustrating an overall configuration of an image processing apparatus according to an embodiment of the present invention. It is a figure which shows typically the principle of selection of the compression process algorithm optimal for the compression process in the image processing apparatus which concerns on embodiment same as the above with a graph. It is a figure which shows typically the principle of selection of the compression process algorithm optimal for the compression process in the image processing apparatus which concerns on embodiment same as the above with a graph. It is an image figure of the image used as the object of processing, and the macroblock formed in the image processing device concerning an embodiment same as the above. It is a flowchart which shows the basic procedure of the process of an image compression in the image processing apparatus which concerns on embodiment same as the above. It is a flowchart which shows the detail of the procedure shown to step S3 in the image processing apparatus which concerns on embodiment same as the above.

  1 to 6 show an embodiment of the present invention.

<Basic configuration>
FIG. 1 is a functional block diagram showing the overall configuration of an image processing apparatus according to an embodiment of the present invention. The image processing system 1A is used for image compression processing.

  The image processing system 1A includes at least one CPU, and includes an image processing device 1, a decoder 2, and various processing units 3 according to the present invention. The image processing apparatus 1, the decoder 2, and the various processing units 3 may each include a CPU.

  The image processing apparatus 1 performs image compression processing.

  The decoder 2 performs a decompression process on the compressed image.

  The various processing units 3 perform various types of image processing, recording, external input / output, and the like. Specifically, a configuration for performing various types of image processing such as image information coordinate conversion, gradation correction, and compression / decompression, a storage medium such as a RAM, an EEPROM, and a hard disk, in which uncompressed image information is temporarily stored. Or a configuration for semi-permanently storing or inputting / outputting image information to / from the image processing apparatus 1, a configuration for inputting / outputting image information to / from external devices, etc. is there.

  As shown in FIG. 1, the image processing apparatus 1 includes an image dividing unit 11 as an “image dividing unit”, an algorithm selecting unit 12 as an “algorithm selecting unit”, and an image compressing unit 13 as an “image compressing unit”. Yes.

  The image dividing unit 11 divides the image into macro blocks.

  The algorithm selection unit 12 recognizes the state of the image for each macroblock and selects an optimal image compression algorithm based on the state of the image.

  The image compression unit 13 performs compression processing in units of macroblocks using the image compression algorithm selected by the algorithm selection unit 12.

<Basic principle of algorithm selection>
The algorithm selection unit 12 compares the numerical values obtained based on the image quality and the code amount when each of a plurality of image compression algorithms is applied to the macroblock currently focused on, and optimizes the one with the smallest or largest numerical value. Select as the image compression algorithm. Hereinafter, the basic principle of the selection will be described.

  2 and 3 are diagrams schematically showing the principle of selection of a compression processing algorithm optimum for the compression processing, using graphs. In the graph 31, the horizontal axis represents the code amount after compression, and the vertical axis represents PSNR (Peak signal-to-noise ratio, hereinafter simply referred to as “PSNR”) that defines the image quality at the time of compression. Show.

For example, as shown in the image diagram of Figure 4, the macroblock ₄₁ 1 of n (n> 1) in the image _40, 41 _2, 41 3, there is · · · 41 _n, the macro block ₄₁ 1, 41 ₂ Let us consider a case where the calculation of the evaluation index and selection of the optimal image compression algorithm have been completed. At this time, the following values (1) and (2) are considered for the macroblocks 41 ₁ and 41 ₂ .

  A point P in FIGS. 2 and 3 shows values obtained based on the values (1) and (2) on the graph 31 and the graph 32.

ただし、
Blocks：評価を開始したマクロブロック４１_１から現在注目しているマクロブロック４１_３までのブロック数

となる。 On the other hand, consider the case where compressed respectively macroblocks 41 ₃ currently focused plurality of image compression algorithms, such as image compression algorithms A and image compression algorithms B. At this time, when the image compression algorithm A is selected and compressed, the average code amount at the time of compression and the mean square error between the image data before compression and the image data reproduced at the time of decompression after compression are expressed by the following formula ( 3) As in (4).

However,
Blocks: The number of blocks from the macro block 41 ₁ that started the evaluation to the macro block 41 ₃ that is currently focused on

It becomes.

  On the other hand, the average code amount and the mean square error when the image compression algorithm B is selected and compressed can be obtained similarly.

  The point A in FIGS. 2 and 3 is shown on the graph 31 and the graph 32 by using values obtained based on the image compression algorithm A based on these points. A point B in FIG. 2 is indicated on the graph 31 and the graph 32 by values obtained based on the image compression algorithm B.

  In general, when compressing an image using an arbitrary image compression algorithm, there is a correlation that the image quality decreases when the compression rate is lowered, and the compression rate increases when the image quality is raised, and a correlation curve may be formed by this relationship. Many. The correlation curve 21 shown in the graph 32 of FIG. 3 schematically shows the correlation between the compression rate of the algorithm and the image quality. That is, the points P, A, and B shown in the graph 32 of FIG. 3 are the same as the points P, A, and B shown in FIG. 2, and the correlation curve 21 is the compression in the image compression algorithm used for calculating the point P. It shows the correlation between rate and image quality.

  For example, taking the point P of the correlation curve 21 as an example, if the compression rate is the same as that of the point P (the position in the horizontal axis direction is the same), the higher the image quality, that is, the higher the evaluation is. In addition, if the image quality is the same as that of the point P (the position in the vertical axis direction is the same), the evaluation is higher when the compression rate is lower, that is, when it is located in the left direction.

  That is, when the point P is used as a reference, the image compression algorithm that is located in the upper left direction illustrated by the arrow 25 in FIG. Here, the further away from the correlation curve 21 indicated by the arrow 25, the higher the evaluation of the image compression algorithm.

  An example of the procedure for selecting the one away from the correlation curve 21 will be described.

In the graph 32 shown in FIG. 3, the tangent line 22 and the contour lines 23 and 24 can be generalized by the following formula (5).
l = c ₁ s + c ₂ (5)
However,
l: code amount s: PSNR value c ₁ : slope of tangent and contour line c ₂ : intersection of contour line and horizontal axis

At this time, as shown in FIG. 3, the value of c ₂ coincides with the intersection of the horizontal axis and the contour lines 23 and 24. Further, as shown in FIG. 3, as the point A and the point B is away from the point P, the value of c ₂ of c ₂ values or contour 24 of the contour 23 becomes small. Therefore, the value of c ₂ of each of the contour lines 23 and 24 to confirm the smaller ones, to the point P, can confirm which of the point A and point B are separated.

を用いると、地点Ａにおいては、

ただし、
Size：ブロック内の画素数
となる。これを変形してｃ_２についての式とすると、

となる。
同様に、地点Ｂにおいては、

となる。
そして、この式（８）のｃ_２の値と式（９）のｃ_２の値とのうち、どちらが小さいかを確認し、最小の値を選択することで、画像圧縮アルゴリズムＡと画像圧縮アルゴリズムＢのうち、最適な画像圧縮アルゴリズムを選択できる。 Here, for example, assuming that the image data is 8 bits (0 to 255), the following PSNR equation (6) with respect to the above equation (5):

Is used at point A,

However,
Size: The number of pixels in the block. When expression for c ₂ by transforming it,

It becomes.
Similarly, at point B,

It becomes.
Then, of the value of c ₂ of the values and formulas c ₂ of the equation (8) (9), to confirm either small, by selecting the minimum value, the image compression algorithms A and image compression algorithms Among B, an optimal image compression algorithm can be selected.

ただし、
Ｓ：評価の指標

ｃ_１：接線２２及等高線２３，２４の傾き

On the other hand, log can be approximated by a straight line, and a constant term is not necessary for comparison. Therefore, for example, the equation shown in the above equation (8) can be simplified as the following equation (10).

However,
S: Evaluation index

c ₁ : inclination of tangent line 22 and contour lines 23 and 24

ただし、
ｃ_１：接線２２及等高線２３，２４の傾き
Blocks：評価を開始したマクロブロック４１_１から現在注目しているマクロブロック４１_３までのブロック数

そして、上記式（１０）と式（１１）とを一般化することにより、以下式（１２）が導出される。
Ｓ＝λＱ＋Ｃ・・・（１２）
ただし、
Ｓ：評価の指標
λ：任意の係数
Ｑ：現在注目しているマクロブロック４１における、圧縮前の画像データと圧縮後展開時に再現された画像データとの２乗誤差
Ｃ：符号量（現在注目しているマクロブロック４１における、圧縮時の平均符号量） From the above, it can be considered that the coefficient in parentheses in the above equation (10) is an analytically optimal value of an arbitrary coefficient λ in the RDO index. Therefore, based on the above equation (10), an arbitrary coefficient λ of the index of RDO can be defined as the following equation (11).

However,
c ₁ : inclination of tangent line 22 and contour lines 23 and 24
Blocks: The number of blocks from the macro block 41 ₁ that started the evaluation to the macro block 41 ₃ that is currently focused on

Then, the following formula (12) is derived by generalizing the formula (10) and the formula (11).
S = λQ + C (12)
However,
S: Index of evaluation λ: Arbitrary coefficient Q: Square error between image data before compression and image data reproduced at the time of decompression in the macroblock 41 currently focused on C: Code amount (currently focused The average code amount at the time of compression in the macro block 41

Therefore, by obtaining the smallest (or largest) evaluation index S in Expression (12), an evaluation equivalent to confirming which point A or point B is away from the point P is performed. Yes. In the image processing apparatus 1 according to this embodiment, the algorithm selection unit 12 calculates an evaluation index S for each image compression algorithm, and selects the one with the smallest value (or the largest value), thereby each macroblock. 41 ₁ , 41 ₂ , 41 ₃ ,..., 41 _n is selected as the optimum image compression algorithm.

<Image compression processing procedure>
FIG. 5 is a flowchart showing a basic procedure of image compression processing in the image processing apparatus 1 according to this embodiment. FIG. 6 is a flowchart showing details of the procedure shown in step S3 in the image compression processing. The processing procedure will be described below with reference to FIG.

First, when an image 40 is input from the various processing units 3 to the image processing apparatus 1 (step S1), the image dividing unit 11 includes n (n> 1) macroblocks 41 ₁ , 41 ₂ , 41. ₃ ,... 41 _n (step S2, image division procedure). 41 ₁ , 41 ₂ , 41 ₃ ,..., 41 _n have substantially the same configuration, and are hereinafter referred to as a macroblock 41 for the sake of simplicity of explanation, unless otherwise required.

  The algorithm selection unit 12 selects an optimal image compression algorithm for each divided macroblock 41 (step S3, algorithm selection procedure).

Macroblock 41 which is currently of interest, if for example, those macroblocks 41 ₁ is first noted ( "No" in step S30), the algorithm selection unit 12, a conventional method, for example, 2 quantized coefficients An optimum evaluation index is calculated by processing such as multiplication, and an optimum compression rate and image quality are obtained (step S31).

On the other hand, if those macroblocks that are currently of interest, for example, macroblock 41 ₃ is focused on the second and subsequent ( "Yes" in step S30), the algorithm selection unit 12, the basic principle of <algorithm selection above The processing is performed based on the method indicated by>.

Specifically, when the algorithm selecting unit 12 sequentially selects from the macroblock 41 ₁ in the upper left and selects the index calculation and optimal algorithms for evaluation, it has focused on the current third macroblocks 41 ₃ think of. In this case, the algorithm selection unit 12 performs processing on the processed macroblocks, for example, the first and second macroblocks 41 ₁ and 41 _{2 in} which the calculation of the evaluation index S and the selection of the optimal algorithm are completed. Then, a comprehensive evaluation index S is calculated (step S32). Specifically, the algorithm selection unit 12 performs an average of image quality (average square error) and code amount when compression processing is performed with an image compression algorithm regarded as optimal for each of the _two macroblocks 41 ₁ and 41 ₂ . An average (average code amount) is obtained. The value of the obtained evaluation index S corresponds to, for example, the point P on the correlation curve 21 in the graph 32 shown in FIG.

Next, the algorithm selection unit 12, for each image compression algorithms, to macroblocks 41 ₃ that are currently of interest to calculate the index S of the evaluation can take in each image compression algorithm (step S33).

  In this embodiment, the algorithm selection unit 12 calculates an index S for each evaluation when each image compression algorithm is applied, based on the above formulas (12) and (11).

When the image processing apparatus 1 has a plurality of image compression algorithms, for example, the image compression algorithm A and the image compression algorithm B, as candidate image compression algorithms, the algorithm selection unit 12 selects these candidate image compression algorithms A and B as the candidate image compression algorithms. respectively used to process the macroblocks 41 ₃ of interest (step S34). That is, the algorithm selection unit 12, first, the image compression unit 13 checks the average code amount C during compression to perform the compression processing of the macroblock 41 ₃ using an image compression algorithm A, then the decoder 2 calculating the image quality Q from the comparison between the macroblock 41 ₃ image data before compression to perform the development process of the image data after expansion to. Next, the algorithm selection unit 12 performs the same procedure using the image compression algorithm B.

Next, the algorithm selection unit 12, shown in FIG. 3, and calculates the inclination _{c 1} of the tangent 22 of the correlation curve 21 at the point P (step S35). In order to obtain the inclination c ₁ , for example, the procedures shown in the following (Method 1) and (Method 2) can be considered.
(Method 1) The value of c ₁ recorded in advance in the algorithm selection unit 12 is used. Specifically, the equation of the correlation curve 21 and a plurality of arbitrary points (not shown) on the correlation curve 21 are recorded in the algorithm selection unit 12 in advance, and the point P and the point P on the correlation curve 21 are recorded. considered a method of an inclination c ₁ the slope of the line connecting the arbitrary position close. By doing so, the value of the slope c ₁ can be quickly obtained with a small amount of calculation.
(Method 2) The algorithm selection unit 12 calculates the value of the slope c ₁ by sequential calculation. Specifically, to generate the values of c ₁ by treating for example the valuated macroblocks in any quality parameter. More specifically, for example, at least one of the following (determination material A) (determination material B) is calculated for the macroblock 41 evaluated immediately before and all the evaluated macroblocks 41, the calculated value is substituted into any evaluation parameters to calculate the value of c ₁ by calculating. In this way, it is possible to determine the value of the slope c ₁ with high accuracy.
(Determination material A) Mean square error between image data at the time of compression and image data reproduced at the time of decompression (determination material B) Average code amount at the time of compression

Further, the slope c ₁ may be obtained by any method other than the above-described method.

Next, the algorithm selection unit 12 uses the average of the mean and the code amount of the calculated image quality, and a slope c ₁ of the tangent 22 which is calculated in step S35 in step S33, based on the above equation (11) An arbitrary coefficient λ is calculated (step S36).

Next, the algorithm selection unit 12, the arbitrary coefficient λ calculated in step S36 are substituted into Equation (12), to calculate the index S of the evaluation of the case of applying the image compression algorithms A macro block 41 ₃ ( Step S37).

The processing from step S33 to step S37 is similarly performed for all the sequentially selected image compression algorithms ("No" in step S38). That is, the algorithm selection unit 12 also performs the processing of step S33 to step S37 for the image compression algorithm B. Thus, the index S of the evaluation of the case of applying the image compression algorithm B to the macroblock 41 ₃ is calculated.

  When the processing of steps S33 to S37 is completed for all candidate image compression algorithms (“Yes” in step S38), the algorithm selection unit 12 determines the evaluation index S calculated in step S32 and the steps S33 to S37. Contrast with the calculated evaluation indexes. Then, an optimal image compression algorithm is selected by selecting an optimal one from these candidate indices (step S39).

Specifically, it calculated in step S33～S37, the case of applying the image compression algorithm B as an index S, and macroblock 41 ₃ rating calculated for the case of applying the image compression algorithms A macro block 41 ₃ Is compared with the evaluation index S calculated. The algorithm selection unit 12 selects the smallest (or largest) of the evaluation indices S as the optimum index, and selects the optimum image compression algorithm based on the optimum index.

  Temporarily, in the process of step S37, the point A is shown on the graph 32 of FIG. 3 based on the value calculated when the image compression algorithm A is applied, and is calculated when the image compression algorithm B is applied. When the point B is indicated on the graph 32 based on the value, the algorithm selection unit 12 selects the image compression algorithm B based on the result.

  After an optimal image compression algorithm is selected for one macroblock 41, if there is a macroblock 41 that has not been noticed yet ("No" in step S4), the algorithm selection unit 12 has not yet paid attention to the macro. Attention is paid to the block 41 in sequence, and the processes of steps S3, S30 to S39 are sequentially performed for them to calculate the evaluation index S and select the optimal image compression algorithm. When the processing is completed for all the macroblocks 41 (“Yes” in step S4), the image compression unit 13 determines each macroblock 41 based on the optimum image compression algorithm selected for each macroblock 41. The image 40 is compressed by compressing the image (step S5, image compression procedure).

As described above, in this embodiment, for each macroblock 41, the slope c ₁ of the tangent line 22 of an arbitrary point in the graph using the image quality and the compression rate when compressed by the respective image compression algorithms as parameters. By selecting the optimal image compression algorithm by using as an index for evaluation, the optimal image compression algorithm in each macroblock 41 is always unambiguous by calculation in a form that always takes into account the state of both image quality and code amount. Can be obtained. Also, simple mathematical formulas are sufficient. Further, by performing such selection sequentially for all the macroblocks 41, an optimal image compression algorithm can be applied to each macroblock 41.

  In this embodiment, numerical values obtained based on image quality and code amount when a plurality of image compression algorithms are respectively applied to the macroblock 41 of interest at present are respectively compared, and the smallest or largest numerical value is compared. By selecting the optimum image compression algorithm, the numerical values obtained by actually applying the respective image compression algorithms to the macroblock 41 can be compared, and the optimum image compression algorithm can be selected based on simple selection of numerical values. .

  In this embodiment, an image compression algorithm that is optimal for compression processing can be selected simply and reliably based on specific calculations.

  In this embodiment, an image compression algorithm optimal for compression processing can be selected with high accuracy from among a plurality of image compression algorithms with a simple calculation.

In the above-described embodiment, the algorithm selection unit 12 applies the smallest one (or the largest one) among the plurality of evaluation indices S calculated by applying the macroblock 41 of interest to a plurality of image compression algorithms. However, the present invention is not limited to this. For example, the axis of the graph using the compression ratio as a parameter shown in FIG. 3 and the above formulas (8) and (9), etc. the size of the intersection c ₂ contour lines 23 and 24 smallest (or largest ones) is selected, may be configured to select an optimal image compression algorithms based on this selection result.

  In addition, the said embodiment is an illustration of this invention, and it cannot be overemphasized that this invention is not meant to be limited only to the said embodiment.

DESCRIPTION OF SYMBOLS 1 ... Image processing apparatus 11 ... Image division part (image division means)
12 ... Algorithm selection unit (algorithm selection means)
13: Image compression unit (image compression means)
41, 41 ₁ , 41 ₂ , 41 ₃ ,... 41 _n ... Macroblock c _1.

In order to achieve the above object, the invention described in claim 1 includes an image dividing unit that divides an image into macro blocks, an algorithm selecting unit that selects an optimal image compression algorithm for each macro block, and the algorithm selection. An image compression unit that performs compression processing for each macroblock using the compression algorithm selected by the unit, wherein the algorithm selection unit includes a plurality of image compression algorithms, and the algorithm The selecting means applies the image quality and the compression rate or the compression rate for the already applied image compression algorithm as the image compression algorithm applied to the macroblock for which the optimal image compression algorithm has already been selected for each macroblock. On the coordinate plane using the corresponding index as a parameter, Forming a correlation curve between the image quality of the image compression algorithm and the compression rate or an index corresponding to the compression rate, and the most recently processed macroblock as the one of the macroblocks to be subjected to the compression processing most recently, An individual corresponding point as a corresponding point between the image quality when compressed using each of at least two of the image compression algorithms and the compression rate or an index corresponding to the compression rate is calculated, and the correlation curve Taking a corresponding point on the correlation curve as a corresponding point when the arbitrary macroblock is compressed by the applied image compression algorithm, and calculating a slope of a tangent of the correlation curve at the corresponding point on the correlation curve, For each individual corresponding point, calculate a contour line that has the same tangent and the same slope and passes through each individual corresponding point. The image compression algorithm for calculating the individual corresponding point located on the contour line that takes an intersection of a height line and a coordinate axis of the coordinate plane and forms a maximum or minimum value among the intersection points, It is characterized in that it is selected as the image compression algorithm most suitable for image compression of the processing macroblock .

The invention according to claim 5 is an image division procedure for dividing an image into macroblocks, an algorithm selection procedure for selecting an optimal image compression algorithm for each macroblock from a plurality of prepared image compression algorithms, An image compression method for performing compression processing for each macroblock using the image compression algorithm selected in the algorithm selection procedure, wherein the algorithm selection procedure includes a step for each macroblock. The coordinates of the applied image compression algorithm as the image compression algorithm applied to the macro block for which the optimal image compression algorithm has already been selected, with the image quality and the compression rate or an index corresponding to the compression rate as parameters On the plane, the image quality of the already applied image compression algorithm and A reduction curve or a correlation curve with an index corresponding to the compression ratio is formed, and the most recently processed macroblock as the one of the macroblocks to be subjected to the compression process is at least two of the plurality of image compression algorithms An individual corresponding point as a corresponding point between the image quality when compressed using each of the two image compression algorithms and a compression rate or an index corresponding to the compression rate, and the arbitrary macroblock on the correlation curve Taking the corresponding point on the correlation curve as the corresponding point when compressed by the applied image compression algorithm and calculating the slope of the tangent of the correlation curve at the corresponding point on the correlation curve, and for each of the individual corresponding points, Contour lines that are equal to the tangent and the slope and pass through the individual corresponding points are calculated, and the contour lines and the coordinate axes of the coordinate plane are calculated. Taking a point, the image compression algorithm calculates the individual corresponding point located on the contour values in the intersection points form the largest or smallest, the optimum wherein the image compression of the immediate processing macro block It is selected as an image compression algorithm.

<Basic principle of algorithm selection>
The algorithm selection unit 12 compares the numerical values obtained based on the image quality when the plurality of image compression algorithms are respectively applied to the currently focused macroblock and the code amount as the “index corresponding to the compression ratio”. The smallest or largest is selected as the optimal image compression algorithm. Hereinafter, the basic principle of the selection will be described.

The point A as the “individual corresponding point” in FIG. 2 and FIG. 3 is shown on the graph 31 and the graph 32 by the values obtained based on the image compression algorithm A based on these points. A point B as an “individual corresponding point” in FIGS . 2 and 3 is shown on the graph 31 and the graph 32 by values obtained based on the image compression algorithm B.

In general, when compressing an image using any image compression algorithm, reducing the compression ratio when rising above image quality, there is a correlation relationship between the compression ratio increases the image quality under want, correlation curve Such relationship is formed There are many cases. The correlation curve 21 shown in the graph 32 of FIG. 3 schematically shows the correlation between the compression rate of the algorithm and the image quality. That is, the points P, A, and B shown in the graph 32 of FIG. 3 are the same as the points P, A, and B shown in FIG. 2, and the correlation curve 21 is the compression in the image compression algorithm used for calculating the point P. It shows the correlation between rate and image quality.

For example, taking the point P as the “corresponding point on the correlation curve” of the correlation curve 21 as an example, if the compression rate is the same as the point P (the position in the horizontal axis direction is the same), the higher image quality, that is, the upper position The evaluation is higher. In addition, if the image quality is the same as that of the point P (the position in the vertical axis direction is the same), the evaluation is higher when the compression rate is lower, that is, when it is located in the left direction.

If the macro block 41 as the “most recently processed macro block” currently being focused on, for example, the macro block 411, is the first _one to be focused on (“No” in step S30), the algorithm selection unit 12 uses the conventional block. An optimal evaluation index is calculated by a method, for example, a process such as squaring of a quantization coefficient, and an optimal compression rate and image quality are obtained (step S31).

On the other hand, when the macro block 41 as the “most recently processed macro block” currently focused on, for example, the macro block 41 ₃ is the second or subsequent one (“Yes” in step S30), the algorithm selection unit 12 The processing is performed based on the method described above in <Basic Principle of Algorithm Selection>.

Specifically, when the algorithm selecting unit 12 sequentially selects from the macroblock 41 ₁ in the upper left and selects the index calculation and optimal algorithms for evaluation, it has focused on the current third macroblocks 41 ₃ think of. In this case, the algorithm selection unit 12 performs processing on the processed macroblocks, for example, the first and second macroblocks 41 ₁ and 41 _{2 in} which the calculation of the evaluation index S and the selection of the optimal algorithm are completed. Then, a comprehensive evaluation index S is calculated (step S32). Specifically, the algorithm selection unit 12 averages the image quality when the compression processing is performed with the image compression algorithm as the “already applied image compression algorithm”, which is regarded as optimal for each of the _two macro blocks 41 ₁ and 41 _2. (Average square error) and average code amount (average code amount) are obtained. The value of the obtained evaluation index S corresponds to, for example, the point P on the correlation curve 21 in the graph 32 shown in FIG.

Claims (5)

Image dividing means for dividing the image into macroblocks;
Algorithm selection means for selecting an optimal image compression algorithm for each macroblock;
An image processing apparatus comprising image compression means for performing compression processing for each macroblock using the compression algorithm selected by the algorithm selection means,
The algorithm selection means comprises a plurality of image compression algorithms, and for each macroblock, the slope of the tangent of an arbitrary point in a graph with the image quality and the compression rate as parameters when compressed by the image compression algorithm. An image processing apparatus characterized by selecting an optimal image compression algorithm by using it as an evaluation index. The algorithm selection means uses, for each macroblock, the image quality after compression and the code amount after compression when compressed by the image compression algorithm in the following equations (1) and (2), 2. The image processing apparatus according to claim 1, wherein the image compression algorithm having the smallest or largest evaluation index value of the expression (1) is selected as the optimum image compression algorithm.
S = λQ + C (1)

However,
S: Evaluation index λ: Arbitrary coefficient Q: Square error C of the image data reproduced in the image data reproduced at the time of decompression with respect to the image data before compression in the macro block currently focused on C: Code amount c ₁ at the time of compression in the macroblock currently focused on: slope of tangent at an arbitrary point of a correlation curve formed by an arbitrary image compression algorithm in a graph using image quality and compression rate as parameters
Blocks: Number of macroblocks sequentially selected for evaluation

The algorithm selection means includes
Previously said algorithm selecting unit wherein c ₁ value in the equation that is recorded (2), and, among the values of the c ₁ in successive Formula calculated by the calculation (2), at least any one The image processing apparatus according to claim 2, wherein the image compression algorithm is selected by using the image compression algorithm. When using the c ₁ in the equation (2) calculated by the sequential calculation, the algorithm selection unit uses the value of the c ₁ generated by processing the processed macroblock with an arbitrary quality parameter. The image processing apparatus according to claim 2, wherein the image processing apparatus is used. Image segmentation procedure for segmenting an image into macroblocks,
An algorithm selection procedure for selecting an optimal image compression algorithm for each macroblock;
An image compression procedure for performing compression processing for each macroblock using the image compression algorithm selected in the algorithm selection procedure,
In the algorithm selection procedure, for each macroblock, the slope of the tangent of an arbitrary point in the graph using the image quality and compression rate when compressed by the image compression algorithm as parameters is used as an evaluation index. An image processing method comprising selecting an optimum image compression algorithm from a plurality of image compression algorithms.

Images