Specular Reflections Removal for Endoscopic Images Based on Improved Criminisi Algorithm

Abstract

Specular reflections always exist in endoscopic images, and they can severely disturb surgeons' observation and judgment. This paper proposes a exemplar-based inpainting algorithm based on the adaptive search range to remove the endoscopic specular reflections. First, a binary thresholding algorithm based on b-channel and morphological dilation operation is used to automatically locate specular reflection regions, and then an improved algorithm based on the Criminisi algorithm is proposed, which uses the method of sequentially repairing highlight regions to reduce the edge contour priority calculation times, and then adopt an adaptive search range strategy to reduce the amount of calculation in the process of searching for the best matching patch, while searching in the additional search area provided by the best frame to improve the accuracy of matching. Experimental results demonstrate that the proposed method can automatically and correctly not only locate but also remove specular reflection regions in endoscopic images. In addition, because the improved method takes much less time than existing methods, and the PSNR and SSIM are higher than existing methods, the specular reflections removal scheme in this paper is better than existing methods.