STANDARD DEVIATION BASED IMPLUSE NOISE FILTER

Abstract

In the field of Digital Image Processing, noise removal plays a very crucial role. The output of any image processing algorithm will depend greatly on the quality of the image sent as input. Many images during acquisition, transmission and processing get corrupted by impulse noise. Thus impulse noise removal becomes an important pre processing step for image processing. Impulse noise usually corrupts only some pixels, leaving a few pixels uncorrupted. Therefore impulse noise removal is normally a two step process. The first step involves classifying the image pixels into corrupted and uncorrupted pixels. The second step deals with restoration of these corrupted pixels. This study introduces a novel approach for impulse noise removal, typically in the range of 10% to 80% noise density. The proposed scheme is a double stage filter, which removes impulse noise based on heuristic calculations of neighboring pixels. In the first stage, “Detector”, the pixels are identified as noisy or noise-free using distance calculations on the neighboring pixels. An adaptive threshold is used to classify these pixels. Once the pixels are identified to be noisy “Filtering” is performed on the noisy pixels. During filtering the image pixels are assigned weight values and the final restoration is done using a weighted median. In order to evaluate the performance of this proposed filter “A Novel Approach for Salt and Pepper Noise Removal based on Heuristic Analysis of Neighboring Pixels” (SPHN), various test images were used. The performance of the SPHN filter is also compared with several other popular techniques. It has been found that not only does the proposed filter work well on a variety of images but also produces results which are significantly better than many popular techniques.