Nested neural networks for image compression

Data compression occurs naturally in the human brain. The brain detects features and the context to any input signal and associates with it a name and form. The use of artificial neural networks for compressing data has been used in the past with some degree of success. The difficulty with this technique is that even though it may achieve a high compression ratio, it provides only the 'approximate' information and loses the 'detail'. In this paper, a new concept has been developed-nested neural networks. These networks are built with a set of networks 'nested' inside the larger network. This scheme has been implemented for data compression of images and the results are promising.