Adaptive Impulse Noise Removal Algorithm using Specific Edge Enhancement Technique for Improved Image Quality

Abstract

The sources of distortion are the motion blurring, distinct types of noises, sensor inadequacy, and error during image transmission. To enhance the performance of image acquisition, transmission, processing and storage system, the visual qualities of theimage have to be maintained and controlled. The main aim is to achieve high Peak Signal to Noise Ratio (PSNR) for restored image by eliminating the noises available in the images and hence, the following techniques are proposed.In order to reduce the complexity of the filter, an Edge Enhanced Non Local Means (EENLM) filter is used. Generally, Neighborhood filters are employed to reduce the image noise by local smoothing of pixels. Initially, the noisy and edge pixels are categorized. Then the edge and noisy pixels are categorized using the Laplacian of Gaussian (LoG) kernel. For suppression of weak and edge pixels, a suitable threshold is applied. The non-local averaging of the input image is performed on the edge and noisy pixels at the final stage so as to improve the system performance. This new technique is performing a Laplacian of Gaussian function along with the non-local means algorithm. The non-local mean method is covering larger area in the image for identifying high similarity instead of taking the neighborhood pixels. Euclidean distance is used to compute the similarity between the pixels in original image. This method provides better PSNR, Structural Similarity Index Measure (SSIM) values and good visual quality when compared to the existing methods.