Hybrid Fuzzy Neural Network Based Still Image Compression

In this paper a hybrid fuzzy neural approach is proposed for the image compression. The inputs to the network are the preprocessed data of original image, while the outputs are reconstructed image data, which are close to the inputs. This process is similar to a typical function approximation problem which involves determining or learning the input-output relations using numeric input-output data. A mutual subset hood based Fuzzy Neural Network is used for image compression application as a function approximator. If the amount of data required to store the hidden unit values and the connection weights to the output layer is less than the original data, compression is achieved. The network is trained for different number of quantization bits with direct impact to compression ratio. It is experimented with different images that have been segmented in the sub image blocks of equal sizes for compression process. The results encourage the possibility of using proposed fuzzy neural network for image compression.