Directional order statistics filtering for suppression of salt and pepper noise

Directional order statistics filtering approaches are performance limited owing to their implementation with larger window sizes (to a maximum size of 9×9) and with the need of iterative operation to filter the residual noise for highly corrupted images. This paper presents a methodology to improve the performance of directional order statistics filters for suppression of salt-and-pepper noise. The proposed filtering approach involves eight directions to identify boundary and non-boundary pixels. If the non-boundary pixels are noise free, they are left unprocessed otherwise; updated by order statistics parameters of the local window. PSNR and SSIM are used as performance parameters to evaluate the quality of the restored images obtained using the proposed approach as well as other related approaches.