| The definition of image compression is to reduce the data, which represents the image. The redundant data is removed without compromising much on the quality of an image. Before storing or transmitting, transformation is applied to the image. By applying a transformation, one can obtain statistically uncorrelated data set. In the reconstruction phase the image is decompressed to get the original image or an image which is as near to original as possible.; The three basic data redundancies are coding redundancy, interpixel redundancy, and the psychovisual redundancy. Most of the compression algorithms work by reducing or eliminating one or more of these data redundancies.; The wavelet transform, transforms the signals into different subbands. The four subbands are LL, LH, HL, and HH. These four subbands represents the sub-images in those subspaces. Once the sub-images has been formed one can use the back propagation algorithm with MLP to find out the corresponding weight matrix and the final subband outputs. |
