Quantization device, method, and recording medium for parallel processing transform coefficients

A quantization method, performed by a quantization device, the quantization method comprising:

• cost calculating, for a plurality of data length candidates, first costs for each of which a degradation of image quality due to compression of and image and a compression ratio of the image are taken into consideration, wherein
• the cost calculating is a common processing partially performed among the plurality of data length candidates, and the calculating the first costs comprising:
• executing the common processing performed among the plurality of data length candidates in parallel, and
• outputting, as an optimal data length, a single data length for which a minimum cost is calculated among the first costs for the plurality of data length candidates;
• calculating second costs for pieces of data in a case of not encoding the pieces of data in parallel, the pieces of data being generated in the compression;
• calculating third costs for the pieces of data in a case of not encoding the pieces of data in parallel,
• calculating, as the first sums, total sums in parallel, wherein
• each of the total sums is a summation of a quantization forward partial sum and a quantization reverse partials sum,
• the quantization forward partial sum is a sum of the second costs for a first part of the pieces of data, the first part being from a first piece to a candidate pieces whose position corresponds to a data length candidate in the plurality of data length candidates, and
• the quantization reverse partial sum is a sum of the third costs for a second part of the pieces of data, the second part being from a piece next to the candidate piece to a last piece.

A non-transitory recording medium that stores a quantization program, when executed by a processor, causes the processor to:

• perform cost calculation processing of calculating, for a plurality of data length candidates, first costs for each of which a degradation of image quality due to compression of the image and a compression ratio of the image are taken into consideration, wherein
• the cost calculation processing is common processing partially performed among the plurality of data length candidates;
• execute the common processing performed among the plurality of data length candidates in parallel;
• acquired, as an optimal data length, a single data length for which a minimum cost is calculated among the first costs for the plurality of data length candidates;
• calculating second costs for pieces of data in a case of not encoding the pieces of data in parallel, the pieces of data being generated in the compression;
• calculating third costs for the pieces of data in a case of not encoding the pieces of data in parallel,
• calculating, as the first sums, total sums in parallel, wherein
• each of the total sums is a summation of a quantization forward partial sum and a quantization reverse partials sum,
• the quantization forward partial sum is a sum of the second costs for a first part of the pieces of data, the first part being from a first piece to a candidate pieces whose position corresponds to a data length candidate in the plurality of data length candidates, and
• the quantization reverse partial sum is a sum of the third costs for a second part of the pieces of data, the second part being from a piece next to the candidate piece to a last piece.