Simultaneous Brightness and Contrast Enhancement Using Derived Inputs and Residual Squeeze Network

In this paper, we propose a method to simultaneously improve the contrast and brightness of low-light images by using the concept of derived inputs and a residual squeeze and excitation block (RSEB) network. The proposed method decomposes the input image into reflectance and illumination based on Retinex theory. Decomposed illumination, gamma corrected illumination, and histogram equalized illumination are combined to create Derived inputs. Then, the RSEB network receives the Derived inputs as input and adaptively rescales the feature map to show the resulting image with enhanced brightness and contrast. Experimental results demonstrates that the proposed method provides excellent result images for both quantitative and qualitative measurements.