Conference paper accepted: Image Compression with Competitive Networks and Pre-fixed Prototypes

The work Image Compression with Competitive Networks and Pre-fixed Prototypes has been published in IFIP International Federation for Information Processing, (247), pp. 339–346.

Abstract:

Image compression techniques have required much attention from the neural networks community for the last years. In this work we intend to develop a new algorithm to perform image compression based on adding some pre-fixed prototypes to those obtained by a competitive neural network. Prototypes are selected to get a better representation of the compressed image, improving the computational time needed to encode the image and decreasing the code-book storage necessities of the standard approach. This new method has been tested with some well-known images and results proved that our proposal outperforms classical methods in terms of maximizing peak-signal-to-noise-ratio values. © 2007 International Federation for Information Processing.

For more details on this work, visit its own page.