Effect of channel errors on source coded image data and the provision of adequate protection of transmission bits

The effects of channel errors on image data coded by means of a zonal transform coder and a threshold transform coder have been studied and experimental results obtained. These form the basis for the selection of a suitable error protection technique. The results have been used in conjunction with five error protection schemes as follows: (i) parameter adjustments only (ii) equal importance protection through Hamming block coding (iii) equal importance protection through Hagelbarger convolutional coding (iv) percentage subimage protection (SIP) using Hamming block codes (v) Percentage subimage protection (SIP) using Hagelbarger convolutional codes. These schemes were selected to provide continuity from one to the next, thus helping in the observation of various trade-offs in the characteristics of the overall transmission systems. A sixth scheme is proposed which turns attention away from the more trditional methods of error protection. This scheme involves maximum likelihood decoding of the Huffman coded image, and thus it can only be applied to codes of the threshold transform type. Although this scheme gives a much improved error protection, its main disadvantage is that it requires considerable data storage and significant search times when decoding the data. The results have been used to deduce the best system for transmission of source coded image data over a channel with a specified error rate and to indicate the preferred technique for the protection of the data.