Zero redundancy error protection of images using self-organizing-maps

This paper proposes the approach of zero redundancy error protection in transmitting images using Self-Organizing Maps (SOM). SOM is a kind of neural network arranging the groups of cluster input having similar characteristics adjacent to each other. SOM is used in this research to design a mapping codebook of gray levels to bit patterns such that similar levels are mapped to similar bit patterns. The codebook designed by SOM for image transmission over noisy channels (Binary Symmetric Channel) are presented and compared with that of using Natural codebook and Graycode codebook. The results indicated that SOM improves the PSNR over both the Natural codebook and Graycode codebook.