Developing a Gaussian Model-based Histogram Equalization Technique for Enhancement of Breast Thermography Images

Abstract

Purpose: Breast cancer is one of the most prevalent diseases among women worldwide. One of the effective ways to reduce the risk of death from breast cancer is early detection by breast screening methods such as thermography. Thermography is non-invasive infrared imaging that detects early symptoms of breast angiogenesis based on the temperature difference and asymmetric patterns between left and right breasts. For better visual perception, it is essential to increase the medical image quality and contrast.
 Materials and Methods: Histogram Equalization (HE) is a common and effective technique for contrast enhancement that uses the whole dynamic range of gray levels. In this paper, we propose to apply the equalization technique to the object part of the image rather than the background. One way is to use Otsu's method for automatic image thresholding. A more efficient approach to extract the body region is to fit a bimodal Gaussian distribution on the temperature information and restrict the equalization on gray level ranges corresponding to temperatures between the mean minus/plus three times of standard deviation.
 Results: We compared the performance of the proposed approach with six conventional HE methods by using objective criteria, including Absolute Mean Brightness Error (AMBE), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSI), and Entropy.
 Conclusion: Based on objective measures, as well as subjective visual inspection of the results, the proposed Gaussian model-based HE has better performance in contrast enhancement and brightness preservation among other methods.