New image compression algorithm using proposed quantization approach

A new quantization method is proposed in this paper. This method is useful for enhancement of compression quality when each kind of neural network is used to compress the image. By quantizing the image with the proposed method, the numbers of samples which must be reconstructed by neural network is reduced. This causes a remarkable increase in quality of the reconstructed image. For testing the proposed method we use autoassociative transform coding and by merging it with the proposed quantization method a new compression algorithm is obtained. Then results of compression by the merged method are compared with some previous works. Obtained results show that the proposed compression algorithm increases the compression quality of the images remarkably. Compression time and complexity in the merged method is also better than JPEG and make it suitable for the systems with low processor and hardware implementation.