A LDCT Image Contrast Enhancement Algorithm Based on Single-Scale Retinex Theory

Image contrast enhancement is a very critical step for automatic medical image processing and analyzing applications. In this paper, we described a novel image enhancement algorithm based on the single-scale Retinex (SSR) theory to enhance the tiny anatomical structures and other regions of interest on the low-dose CT (LDCT) images. This algorithm applies a three-stage approach: (a) separating the input images into illumination component and reflectance component, and calculating the approximate luminance of the original images; (b) normalizing the illumination component for the image dynamic range adaptively and manipulating the reflectance component with a nonlinear function; and (c) Finally, combining the normalized illuminance with the Gamma corrected reflectance to obtain the results. We have tested the proposed algorithm on 2 CT data sets. Comparing with the conventional histogram equalization, the Frankle-McCann Retinex algorithm and McCann99 algorithm, the proposed algorithm yields better performance for LDCT images enhancement.