Impulse Noise Detection and Removal Using First Order and Second Order Difference with Multistage Directional Statistics

Most of the image processing techniques does not perform well when images are corrupted with noise therefore restoration of corrupted images is considered as preprocessing step in most of the image processing applications. Detection and removal of impulse noise is an active area of research and in this paper a novel technique has been proposed which remove Random value impulse noise (RVIN) from digital gray scale images. The proposed method consists of two main modules. First module uses directional based statistics to detect noisy pixels. Detected noisy pixels are passed to the filtering scheme for estimation of their noise free values. Proposed technique has been compared using well known performance measure peak-signal to-noise ratio. Simulations show that the proposed filter can provide excellent performance of suppressing impulse noise in all situations.