Color Image Compression Using Block-Based Independent Component Analysis

A new approach for image compression has been presented, which transform the image data by block-based independent component analysis (ICA). Proposed method uses Hyvarinen and Oja´s fast independent component analysis (FastICA) algorithm followed by arithmetic coding to perform compression. The number of required independent components are determined by measuring the quality parameters such as signal-to-noise ratio or root mean squared error (RMSE). A quantitative as well as qualitative comparision has been done between compression using ICA and PCA. The proposed FastICA -based compression is found more effective than PCA -based compression for image data.