An adaptive switching median filter with anisotropic linking PCNN noise detection for salt and pepper noise reduction

This paper proposes a switching scheme for salt and pepper noise reduction by combining a noise detection algorithm based on a simplified pulse coupled neural network (PCNN) with a simple adaptive median filter. The simplified PCNN utilizes an adaptive synaptic weight matrix created by anisotropic linking mechanism to achieve anisotropic linking model, that is the interconnections between neurons with large absolute difference in intensity will be interrupted. Therefore, the neurons corresponding to noise corrupted pixels will receive smaller feedback signal from the neighborhood and generate smaller internal activities compare with the ones corresponding to noise free pixels. The impulse will be detected by setting an appropriate dynamic threshold. After the PCNN based noise detection scheme, the pixels contaminated by salt and pepper noise will be restored by a simple adaptive median filter. Experimental results prove that the proposed switching median filter outperform over the conventional methods in both noise reduction and detail preserving.