Edge preserving restoration of random valued impulse noises (EPRRVIN)

Abstract

To remove random valued impulse noise from the digital images, a novel two step method has been proposed. The first step is to classify whether the center pixel in the 5 × 5 window is noisy or not, which is done using all neighbor directional weighted pixels in a 5 × 5 mask. The proposed algorithm performs simple arithmetic absolute differences on the pixels aligned in the four main directions with the center pixel. On detection of noisy pixels an advance median filter has been proposed for restoration of noisy pixel where a variable window consisting of 3 × 3 and 5 × 5 respectively have been used. To optimize the results, a 3-D search space has been used on user supplied parameters in a wide range. Proposed algorithm obtains better results in suppression of RVIN in digital images than the existing operators. The algorithm also preserves fine textures and details in the restored images.