Color Image Compression with Modified Forward-Only Counterpropagation Neural Network: Improvement of the Quality using Different Distance Measures

A modified forward-only counterpropagation neural network (MFO-CPN) for color image compression is proposed. It uses several higher-order distance measures for calculating winning node. It also incorporates nonlinear adjustment of learning rates in both the layers. Results with these distance functions are compared. Proposed modifications leads to improvement in the image quality and faster convergence of network.