Skip to Main Content
This paper discusses the use of self-organizing map (SOM) in the image coding. The image is partitioned into blocks each of which is fed to the SOM as the training vectors. Then they are quantized into smaller number of vectors that is used as codeword to reconstruct the image. Thus the data size is reduced. In reconstructing the image, each block is filled with the corresponding codeword. The feasibility of the system is tested by computer simulations and the effect of the sizes of blocks and SOM on the compression rate and image quality is studied. Then the SOM based hardware image coding system is designed, and its functionality is also tested. It is estimated that the hardware SOM can code a single image within 1.2 ms.