A Parameter Optimized, Automatic Threshold Selection for Enhancing Edge Preserving Denoising and Segmentation

Abstract: Image denoising is the most challenging issue in the field of image processing. The main task experienced during image denoising process is to categorize the components of original image from the noisy image. Hence, efficient segmentation mechanisms are needed for categorizing the noise. In Image Denoising With Edge-Preserving and Segmentation Based on Mask Non Harmonic Analysis (ID-EPS), image denoising was done in spatial domain by preserving the edges with fuzzy boundaries. The information in edges was preserved from the input image by edge detection and segmentation with the help of fixed threshold and segmentation parameters values. But in general, Performance of denoising depends on the values of these parameters. Hence optimization of threshold and parameter values is required. In proposed Enhanced Image denoising with Edge Preserving segmentation (EID-EPS) method, performance of image denoising and segmentation result is improved by using frequency domain coefficients like Discrete Cosine Transform (DCT) which segments the similar frequency patches from the input image. Moreover, Iterative Threshold Selection method is introduced for automatically selecting the threshold value for every successive iteration. Parameter values are also optimized by using Servo Parameters Optimization Algorithm in which values of parameters are selected from the input images based on the target image. Experiments are conducted to analyze the performance of denoising with the existing system.

Keywords: Image denoising; Mask Non Harmonic Analysis; Segmentation; Frequency Domain coefficients; Optimization Algorithm.