Advanced Framework for Effective Denoising the Enhanced Thermal Breast Image

Abstract

Nullifying the noise and redundancy in the breast thermogram is still a critical challenge. In this paper, a framework is designed for the purpose of classification of noisy RGB thermal breast images. The proposed framework includes enhancement in frequency domain, advanced mixed denoising and color correction for thermal images. The enhancement of thermogram uses curvelet transform that is applied to the V component of the HSV derived from RGB thermogram. Gain-controlled bihistogram equalization operation is performed to the detailed component of V to enhance the image quality. The advanced mixed denoising algorithm is also presented using a combination of Gaussian and bilateral filter for denoising the enhanced thermogram. Spatially varying color correction (SVCC) technique is applied, which is based on an optimum linear color correction matrix that is calculated from the local blocks of enhanced image. The classification outcome of this proposed framework is more encouraging compared with the results of the existing methods for thermogram classification.