Abstract: We present a recent simulation of our neural net model for image compression (SFSN) which is based on the Kohonen SOFM system. Our previous work was limited to a certain scope of image' domains. Our updated simulator is meant to be very general via a well constructed universal codebook for each domain of images. It shows an improvement over the traditional peer non-neural models (e.g., Wavelet and JPEG) in some image domains. In this paper, we present our neural compression simulator and our most recent results in some important domains, such as satellite and documents imaging.