Presentation | 2013-04-25 Designing and Evaluation of CNN Predictors based on Weighted Feature Quantity and Estimate of Coding Bits Keisuke TAKIZAWA, Hisashi AOMORI, Tsuyoshi OTAKE, Ichiro MATSUDA, Susumu ITOH, Mamoru TANAKA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In recent years, high efficiency image coding schemes become indispensable to archive very high quality images for art works and medical diagnosis because of development of digital imaging devices. Owing to this background, the hierarchical lossless image coding algorithm using CNN was proposed in our previous work. In this paper, we adopted it as the basic method. In the basic method, six types of templates are used for dealing with the local structure of an image, and the CNN parameters are decided to minimize the entropy of prediction residuals in each coding step called stage. However, this parameter determination algorithm is not global optimization but local optimization. The actual coding rate is given by the arithmetic coding using context-adaptive probability density functions (PDFs). In this paper, to deal with the these difficulties, the coding parameter learning processes are started from the last level, and the coded prediction residuals data are transmitted to the previous level. This backward propagation makes it possible to learn optimal parameters with respect to the actual coding rate. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | discrete-time cellular neural network / estimate of coding bits / particle swarm optimization / supervised learning |
Paper # | NLP2013-1 |
Date of Issue |
Conference Information | |
Committee | NLP |
---|---|
Conference Date | 2013/4/18(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Nonlinear Problems (NLP) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Designing and Evaluation of CNN Predictors based on Weighted Feature Quantity and Estimate of Coding Bits |
Sub Title (in English) | |
Keyword(1) | discrete-time cellular neural network |
Keyword(2) | estimate of coding bits |
Keyword(3) | particle swarm optimization |
Keyword(4) | supervised learning |
1st Author's Name | Keisuke TAKIZAWA |
1st Author's Affiliation | Faculty of Science and Technology, Tokyo University of Science() |
2nd Author's Name | Hisashi AOMORI |
2nd Author's Affiliation | Faculty of Engineering, Chukyo University |
3rd Author's Name | Tsuyoshi OTAKE |
3rd Author's Affiliation | Faculty of Engineering, Tamagawa University |
4th Author's Name | Ichiro MATSUDA |
4th Author's Affiliation | Faculty of Science and Technology, Tokyo University of Science |
5th Author's Name | Susumu ITOH |
5th Author's Affiliation | Faculty of Science and Technology, Tokyo University of Science |
6th Author's Name | Mamoru TANAKA |
6th Author's Affiliation | Faculty of Science and Technology, Sophia University |
Date | 2013-04-25 |
Paper # | NLP2013-1 |
Volume (vol) | vol.113 |
Number (no) | 15 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |