Fusion-Based Backlit Image Enhancement Using Multiple S-Type Transformations For Convex Combination Coefficients

Abstract

The backlit image contains both dark and bright regions, and the visibility of such regions is low. It is important to improve the visibility of such images for object recognition. In general, it is difficult to improve the local contrast of back-lit images by using a global contrast enhancement method based on tone mapping because the intensity histogram is severely distorted. On the other hand, local contrast enhancement improves the visibility in the local regions by enhancing local intensity differences. However, the image may be transformed into an unnatural image because of over-enhancement or disordered lightness order. In this paper, we propose a fusion-based backlit image enhancement method. Concretely, we perform image enhancement using various S-type curves for convex combination coefficients to produce multiple enhancement results. By combining enhanced images, it is possible to obtain global and local enhanced images. Through the experiments, we showed that the proposed method produces global and local enhanced images giving natural impression.