TDCCN: A Two-Phase Deep Color Correction Network for Traditional Chinese Medicine Tongue Images

Abstract

In Traditional Chinese Medicine (TCM), tongue diagnosis is an indispensable diagnostic method. Due to the limitations of acquisition devices and variations of the illumination, there is a significant color distortion between the captured tongue image and the human visual perceived image. In this paper, we proposed a Two-phase Deep Color Correction Network (TDCCN) for TCM tongue images. In the first phase, a deep color correction network was designed to model the mapping between the captured image and the target objective chromatic values under a standard lighting condition. The first phase provides consistent color tongue images by using different cameras and capture devices. The output tongue images at this phase can be used for further automatic quantitative analysis. The second phase provides flexible color adjusting options to adapt to different work environments and the subjective preference of doctors for visually perceived color appearance. Only three additional parameters are used to describe the adjustment operation. Experimental results show that our method achieved state-of-the-art performance in objective color correction and obtained satisfactory perceptual adaptation.