Block based Kalman filter algorithm for blind image separation using sparsity measure

This paper presents blind image separation using Kalman filter algorithm. Blind image separation is concerned with recovering the original source images given the sources being mixed with an unknown medium. Kalman filter provides optimal recursive solution to the estimation of the unknown mixture. A novel method called block Kalman filter is used for effectively extracting the images from the mixed images. The observed image is transformed into sparse blocks and the best block is selected using sparsity measure ℓ0 norm as a cost function. Experimental results suggest that the proposed method provides significant separation compared to the Infomax algorithm. Performance evaluation results are presented.