Design and methodology of algorithm for the enhancement of breast tumor images

Abstract

Breast cancer is the second largest disease in the world having high mortality rate among ladies. Breast image enhancement is required for the foundation of high accuracy segmentation. Several features are extracted from the ROI of mammogram in order to detect the cancer tumor in the early stage. Ultrasound breast images have the speckle noise during the capturing of image. In mammogram images also, the noises like Guassian, Normal and Poisson are introduced. There are no proper enhancements algorithms are designed to enhance the images for the preprocessing of the images. The image enhancement is broadly categorized into two different domains such as spatial and frequency. The spatial domain algorithms will smoothen the image quality and frequency domain algorithms will sharpen the image. Till today there is no concrete work on suitable algorithm methods using different domains of enhancement. The research work has been implemented in two stages. In the first stage we are identifying various noises in the mammogram and ultrasound images. In second stage the algorithms like mean, median, Homomorphic, Weiner, Modified Gabor, Anisoid diffusion filter, Contrast limited Adaptive Histogram equalization (CLAHE), Bilateral filter and Guided filter are worked out. Among all these filters we modified Gabor filter that clearly highlights the portion of the tumor. For all these filters Mean Square Error (MSE), RMSE, PSNR, VAR, SSIM, SD are computed. The best PSNR is obtained for the modified Gabor filter.