An enhanced non-local variational level set segmentation and bias correction
Chandradatta Verma, Chandrahas Sahu · 2012
Image segmentation is an initial and vital step in a series of processes aimed at overall image understanding. Noise and Irregularity in light intensities are the major bottleneck in the segmentation process which generally present in real images. Most of the segmentation processes are region based process and depends on regularity of intensities in that region, which lead to the faulty segmentation of images that are affected by noise and intensity inhomogenity. This paper presents a novel approach for segmentation of the images with irregular intensities and noise. Non-local denoising models provide excellent results because these models can denoise smooth regions or/and textured regions simultaneously, unlike standard denoising models. We presented a integrated model which correct the image as well as segment the image. A non-local denoising algorithm presented which denoise the image in the preprocessing step. Following that Local clustering criteria function is presented using K-means clustering algorithm for the images with irregular intensities. This local clustering criteria function when formulated in the level set, it gives better segmentation model. Continuous global minimization of energy function will give segmentation result and bias field which is a cause of irregular intensities in image. Bias corrected image can be obtained by removing obtained bias field form corrupted image. Therefore our method segments the image and at the same time corrects the effect of irregular intensities and noise. A MATLAB code has been implemented based on this method and it gives good results in all cases including the presence of irregular intensities in image and other non-local noise. Our method also has better performance characteristics like robustness and accuracy compared to previous segmentation techniques.