A method for selective color images compression

Some applications that make use of digital images do not allow that the last are compressed with loss of information. A solution for this problem can be the use of a selective compression method, where determined regions of the image where the loss of information is not tolerable (regions of interest RoI) are compressed without losses and the remaining portion with losses, making possible to reach high compression ratios. This paper proposes a selective compression method of digital color images where the perfect reconstruction of the regions of interest is guaranteed. The method uses competitive neural networks, mathematical morphology, Huffman adaptive coding and wavelet transform.