JP2009273064A - Encoder evaluation method, program, and encoder evaluation device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To evaluate whether encoded data are encoded by an optimum compression method or not. <P>SOLUTION: This encoder evaluation method is executed by this encoder evaluation device 1 having a processor 2, and capable of accessing to an original image data storage part 3 for storing original image data therein, and an encoded data storage part 4 for storing encoded data obtained by encoding the original image data therein. The encoder evaluation method includes processes of: reading the original image data from the original data storage part 3 by the processor 2; generating predicted image data from the original image data by a plurality of prediction methods to calculate cost on a prediction method basis; reading the encoded data from the encoded data storage part 4; and evaluating cost calculated from the original image data and cost calculated from a parameter of the encoded data. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an encoder evaluation method, a program, and an encoder evaluation device.

Since the data size of moving image data is very large, various compression methods have been conventionally devised and standardized. Representative examples include MPEG-2, MPEG-4 Visual, H.264. H.264 / MPEG-4 AVC. Along with this, various devices and software (encoders) for compressing original moving image data (original image data) in accordance with a target compression method have been developed. Even in an encoder that compresses to the H.264 / MPEG-4 AVC format, the compression rate, the image quality after compression, and the like differ depending on the developed manufacturer. For this reason, when using an encoder, it is necessary to select an encoder suitable for the purpose, and it is necessary to evaluate the encoder. In general, the encoder is evaluated by comparing original image data and data restored (decoded) from data compressed by the encoder (see, for example, Patent Document 1).
Japanese Patent No. 3529305

  However, with the conventional method, it is possible to determine how much the restored data reproduces the original image data, but it is not possible to determine whether or not an optimal compression method has been selected during encoding. was there.

  The present invention has been made in view of such a problem, and provides an encoder evaluation method, a program, and an encoder evaluation apparatus capable of evaluating whether encoded data is encoded by an optimal compression method. For the purpose.

In order to solve the above problems, an encoder evaluation method according to the present invention includes a processing device, and stores an original image data storage unit in which original image data is stored and encoded data in which the original image data is encoded. An encoder evaluation method executed by a computer that can access an encoded data storage unit, wherein a processing device reads original image data from the original image data storage unit, and predicts image data by a plurality of prediction methods from the original image data And calculating the cost for each prediction method, the process of reading the encoded data from the encoded data storage unit, and the cost calculated from the original image data as C _0, and calculating the cost calculated from the parameters of the encoded data As C _e , the following formula d = C ₀ −C _e
And calculating the evaluation value d.

In such an encoder evaluation method, the prediction method is md, the original image data is f, the prediction image data generated by the prediction method md is g (md), and the distance value between the original image data and the encoded data is D. (F, g (md)), and the code amount in the prediction method md is R (md), the cost cost (md) in the prediction method md is expressed by the following formula: cost (md) = D (f, g (md) )) + Λ × R (md)
However, λ is preferably calculated by a constant.

  A program for causing a computer to execute the encoder evaluation method according to the present invention can be created. This program can be stored in a storage medium such as a flexible disk, a CD-ROM, a magneto-optical disk, a semiconductor memory, a hard disk, or the like. Stored in the device. In some cases, it may be distributed as a digital signal via a network or the like. At this time, intermediate processing results are temporarily stored in a storage device such as a main memory.

An encoder evaluation apparatus according to the present invention includes an original image data storage unit that stores original image data, an encode data storage unit that stores encoded data obtained by encoding the original image data, and an original image data storage unit An original image data reading unit that reads the original image data from, a cost calculation unit that generates predicted image data from the original image data by a plurality of prediction methods, calculates a cost for each prediction method, and encode data from the encoded data storage unit The following formula d = C ₀ −C _e , where C ₀ is the cost calculated from the encoded data reading unit to be read and the original image, and C _e is the cost calculated from the parameters of the encoded data.
And an evaluation unit for calculating an evaluation value d.

In such an encoder evaluation apparatus, the prediction unit sets the prediction method to md, the original image data to f, the prediction image data generated by the prediction method md to g (md), and the original image data and the encoded data. When the distance value is D (f, g (md)) and the code amount in the prediction method md is R (md), the cost cost (md) in this prediction method md is expressed by the following equation: cost (md) = D (f , G (md)) + λ × R (md)
However, λ is preferably calculated by a constant.

  When the encoder evaluation method, the program, and the encoder evaluation device according to the present invention are configured as described above, it is possible to evaluate whether or not the encoded data is encoded by an optimal compression method.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. First, the configuration of an encoder evaluation apparatus that evaluates an encoder that encodes encoded data from original image data and encoded data will be described with reference to FIG. The encoder evaluation device 1 includes a processing device 2 that performs evaluation processing, an original image data storage unit 3 that stores original image data, and encoded data that stores encoded data that is a result of encoding the original image data. The storage unit 4 includes a display device 5 that outputs original image data, encoded data, and evaluation results, and an input device 6 that gives instructions for starting evaluation and displaying evaluation results. The processing device 2 is composed of a computer having a CPU, a main memory, etc., the original image data storage unit 3 and the encoded data storage unit 4 are composed of an external storage device such as a hard disk, and the display unit 5 is a display or the like. The input unit 6 includes a keyboard and a mouse.

  The processing device 2 includes an original image data reading unit 7 that reads original image data from the original image data storage unit 3, a cost calculation unit 8 that calculates a cost in encoding from the original image data, and an encoding data storage unit 4 that encodes The encoded data reading unit 9 that reads data, the evaluation unit 10 that compares the cost calculation result by the cost calculating unit 8 with the cost calculated from the parameters of the encoded data, and the original image data reading unit 7 read the data. The original image display unit 11 that outputs the original image data to the display device 5, the decoding unit 12 that decodes and restores the encoded data read by the encoded data reading unit 9, and the image data that is restored by the decoding unit 12 Is output to the display device 5, and the result evaluated by the evaluation unit 10 is displayed on the display device 5. Composed evaluation result display unit 14 for outputting. Note that these functions are implemented as a program executed by the processing device 2, for example.

  The encoder evaluation process executed by the encoder evaluation apparatus 1 will be described with reference to FIG. It is assumed that the original image data and the encoded data that is the result of encoding the original image data are stored in the original image data storage unit 3 and the encoded data storage unit 4 in advance. The encoder evaluation apparatus 1 first reads original image data from the original image data storage unit 3 by the original image data reading unit 7 (step S100). The read original image data is transferred to the original image display unit 11 and the cost calculation unit 8. The reading of the original image data may be configured to read the entire original image data at once, or may be configured to specify and read a predetermined frame according to an instruction from the input device 6 ( In the following description, a case where a predetermined frame is designated and this frame is read and evaluated will be described).

  Next, the cost calculation unit 8 of the encoder evaluation apparatus 1 applies various compression methods to one frame of the original image data read in step S100 described above, and calculates the cost by each compression method (step). S200).

  Here, referring to FIG. An encoder process for compressing (encoding) moving image data in the H.246 / MPEG-4 AVC format will be described. In this method, moving image data is read frame by frame and compressed. First, one frame of original image data is read from a storage unit (hard disk or the like) (step S600). Then, it is determined whether this frame corresponds to an I picture, a P picture, or a B picture (step S610). If it is an I picture, intra prediction processing is executed for this frame (step S620), and the P picture Alternatively, in the case of a B picture, an inter prediction process is executed for this frame (step S630). In these prediction processes, a prediction image by each method is created, cost calculation is performed with the original image, and a prediction image having the minimum cost is determined. Based on these prediction results, a DCT (Discrete Cosine Transform) calculation is performed on this frame to determine a quantization count (step S640), and entropy encoding is further performed to generate encoded data (step S640). In step S650, the result (encoded data) is written in the storage unit (step S660), and the process for one frame is completed. Then, the processing from step S600 to step S660 is executed for all frames (step S670), and the encoding of the original image data is finished.

  In this way, the encoder determines the least costly compression (encoding) method by creating a predicted image from the original image data (frame) and calculating the cost. However, the encoding process is performed with a limited compression time. In order to perform this, not all compression methods can be evaluated, and the optimum compression method is not always applied. Therefore, the cost calculation unit 8 of the encoding evaluation apparatus 1 is configured to calculate each cost by a compression method (prediction mode) set in advance for the original image data (frame).

  Here, H. In the case of 246 High Profile, it is necessary to determine any one mode from four types of 16 × 16 pixel intra prediction modes, nine types of 4 × 4 pixel intra prediction modes, and nine types of 8 × 8 pixel intra prediction modes. Yes, the cost calculation unit 8 calculates costs in all these modes. For example, in the case of the 4 × 4 pixel intra prediction mode, the entire frame is divided into 4 × 4 pixel images based on the intra prediction mode, and cost calculation described later is performed for each 4 × 4 pixel. The mode when the value is the smallest for each 4 × 4 pixels is adopted.

  There are SAD (Sum of Absolute Differences), SSD (Sum of Squared Differences), SATD (Sum of Absolute Transformed Differences), etc. as quantities representing the difference between the original image and the predicted image. Specifically, SAD and SSD are It is expressed as the following formulas (1) and (2). Here, f (x, y) is an original image of M × N pixels, g (x, y) is a predicted image generated in a certain mode, and x and y are positions of the respective pixels.

  When the Hadamard matrix is H, the SATD is expressed as the following equation (3). However, the Hadamard matrix H in the case of 4 × 4 pixels is expressed as the following equation (4).

  Since the method used to calculate the distance values such as SAD, SSD, and SATD described above varies depending on the encoder designer, the cost cost in a certain mode md is generally expressed by D (f, g). (Md) is expressed as the following equation (5). Note that the pixel positions x and y are omitted, and the predicted image in the mode md is g (md). In equation (5), R (md) is a code amount in a certain mode md, and λ is a constant.

cost (md) = D (f, g (md)) + λ × R (md) (5)

  The cost calculation unit 8 calculates the cost of the 4 × 4 pixel intra prediction mode for the frame read out in step S100 by the above-described method, and further calculates the 16 × 16 pixel intra prediction mode and the 8 × The same cost calculation is performed in the 8-pixel intra prediction mode. Here, the result of the cost calculation is stored in a buffer memory or the like that temporarily stores information, although not shown in FIG.

When the cost calculation of the read original image data of one frame is performed as described above, the encoder evaluation device 1 next reads the encoded data of the corresponding frame from the encoded data storage unit 4, and the evaluation unit 10 and the decoding unit 12 (step S300). Then, for each block described above (for example, for each 4 × 4 block), the evaluation unit 10 compares the cost calculated from the original image data with the cost calculated from the parameters of the encoded data, and the encoded data. Is the best one (step S400). In this evaluation, for example, the cost calculated from the original image data (the lowest cost among the costs of each mode calculated in step S200 described above) is C _0, and the cost calculated from the parameters of the encoded data is C _e. When the evaluation value d shown in the following equation (6) is closer to 0, it is judged as the best.

d = C ₀ −C _e (6)

  As described above, the above-described evaluation value d is calculated for each block of the designated frame of the original image data, and the result is displayed on the display device 5 by the display result display unit 14 to display one frame. The process for is terminated (step S500).

  As a method for displaying the evaluation results, as shown in FIG. 4, the original image screen G1 displayed by the original image display unit 11 and the encode screen displayed by the encode image display unit 13 are displayed on the display device 5. By simultaneously displaying G2 and the evaluation result screen G3 displayed by the evaluation result display unit 14, the comparison between the original image data and the encoded data can be facilitated. Here, the original image display unit 11 displays the entire designated frame on the original image screen G1, and 16 × 16 pixels are defined as one macro block MB, and a grid-like boundary line indicating this block is used as the original image. It is configured to be displayed in a superimposed manner. The encoded image display unit 13 is configured to display the encoded data corresponding to any one of the macro blocks MB specified by the input device 6 as the encoded screen G2 in the original image screen G1 described above. In FIG. 4, since 4 × 4 pixel intra prediction is employed, the 16 × 16 pixel region is divided into 4 × 4 regions as the encoded image G2a displayed on the encode screen G2. (One of the divided areas corresponds to 4 × 4 pixels). Furthermore, the evaluation result display unit 14 is configured to display the evaluation result (cost calculated for each mode) in the macro block MB displayed on the encode screen G2 on the evaluation result screen G3. Also in the evaluation result screen G3 shown in FIG. 4, the 16 × 16 pixel macroblock MB is divided into 4 × 4 4 × 4 pixel blocks, and the calculation result for each mode for each of these blocks. Is displayed.

  In the result shown in FIG. 4, in the encoded data (encoded image), mode 4 is selected from 4 × 4 pixel intra prediction in block (0, 0) in macro block (21, 8). . Further, SSD is selected as the distance calculation. On the other hand, looking at the cost calculation result in the same block, it can be seen that mode 4 shows the smallest SSD value, and as a result, the encoder has selected the optimal compression method in this block. I understand that.

  In the above description, the case has been described where the cost is calculated and evaluated from the distance value between the original image and the predicted image. However, the present invention is not limited to this, for example, intra prediction. It is also possible to calculate and evaluate the cost in determining the mode (Y component, chroma component), determining partition partitioning in inter prediction, determining a motion vector in inter prediction, selecting a reference image in inter prediction, and the like. Further, in the above description, a case has been described where one frame is selected from the input device 6 and the above-described evaluation process is performed. However, the entire original image data is read and all the images are read. It is also possible to perform the above-described processing on the frames and store them in a buffer memory or the like, and designate the frames one by one from the buffer memory and display the evaluation results on the display device 5.

  Furthermore, in the evaluation results shown in FIG. 4, the comparison when the original image and the predicted image are compression methods of 4 × 4 pixels has been described, but it is also possible to compare the original image and the predicted image having different numbers of pixels. is there. For example, when comparing 16 × 16 pixels with 4 × 4 pixels, the costs of 4 × 4 4 × 4 pixels corresponding to the region of 16 × 16 pixels are totaled, and the total value is calculated with 16 × 16 pixels. It is possible to evaluate by comparing the costs.

It is a block diagram which shows the structure of an encoder evaluation apparatus. It is a flowchart which shows the process of an encoder evaluation method. H. 3 is a flowchart showing processing when encoding moving image data in the H.264 / MPEG-4 AVC format. It is explanatory drawing which shows the example of a display to a display apparatus, Comprising: (a) shows an original image screen and an encoding screen, (b) shows an evaluation result screen.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Encoder evaluation apparatus 2 Processing apparatus 3 Original image data storage part 4 Encoded data storage part 7 Original image data reading part 8 Cost calculation part 9 Encoded data reading part 10 Evaluation part

Claims (5)

An encoder evaluation method executed by a computer having a processing device and accessible to an original image data storage unit storing original image data and an encoded data storage unit storing encoded data obtained by encoding the original image data There,
By the processing device,
A process of reading the original image data from the original image data storage unit;
Generating predicted image data from the original image data by a plurality of prediction methods, and calculating a cost for each prediction method;
A process of reading the encoded data from the encoded data storage unit;
Assuming that the cost calculated from the original image data is C ₀ and the cost calculated from the parameters of the encoded data is C _e , the following expression d = C ₀ −C _e
A process of calculating an evaluation value d by
An encoder evaluation method including: The prediction method is md, the original image data is f, the prediction image data generated by the prediction method md is g (md), and the distance value between the original image data and the encoded data is D (f , G (md)) and the code amount in the prediction method md is R (md), the cost cost (md) in the prediction method md is expressed by the following equation: cost (md) = D (f, g (md) )) + Λ × R (md)
However, the encoder evaluation method according to claim 1, wherein λ is calculated by a constant.   A program for causing a computer to execute the encoder evaluation method according to claim 1. An original image data storage unit for storing original image data;
An encoded data storage unit that stores encoded data obtained by encoding the original image data;
An original image data reading unit for reading the original image data from the original image data storage unit;
A cost calculator that generates predicted image data from the original image data by a plurality of prediction methods, and calculates a cost for each of the prediction methods;
An encoded data reading unit that reads the encoded data from the encoded data storage unit;
Assuming that the cost calculated from the original image data is C ₀ and the cost calculated from the parameters of the encoded data is C _e , the following expression d = C ₀ −C _e
And an evaluation unit that calculates an evaluation value d. The prediction unit
The prediction method is md, the original image data is f, the prediction image data generated by the prediction method md is g (md), and the distance value between the original image data and the encoded data is D (f , G (md)) and the code amount in the prediction method md is R (md), the cost cost (md) in the prediction method md is expressed by the following equation: cost (md) = D (f, g (md) )) + Λ × R (md)
However, the encoder evaluation apparatus according to claim 4, wherein λ is calculated by a constant.