HOS-based filtering scheme for stereo image compression

In this paper, a new HOS-based filtering scheme for stereo image compression applications is introduced. This method removes the effects of noise in a stereo image pair by applying the left image as the reference input to a 2-D transversal filter while the right image is used as the desired output. The filter weights are computed using a block-based cumulants matching method. Using higher order statistics is appropriate in case where the stereo image pair is corrupted by zero-mean additive noise. Experiments were performed on a test and real stereo image pairs and results were compared with the least square (LS) and block-matching based methods.