Fuzzy modified forward-only counter propagation network to improve image compression

Images play an important role in the making of attractive websites. And because of this it is impossible to remove images from the websites. Due to high costs of bandwidths and time consumption for up loading and downloading, image compression has become a necessity part of web services. In this paper, a Modified Forward-Only Counter Propagation (MFOCPN) neural network is suggested, in which the model inherits general space localization and frequency correlation of wavelets. The proposed neural network includes two training phases in the first phase fuzzy concepts is applied. in this phase, instead of winning a cluster, the weight of every cluster is updated regarding to the rate of similarity between the input vectors. The Lena image in different sizes has used to evaluate the suggested method. The experimental results in most cases has shown that the returned image in this method is comparison to Wavelet-MFOCPN has more quality of image compression. We achieved on average (%10) less MSE, based on a comparison between original image and returned of compression.