Visually aesthetic image contrast enhancement

摘要

The existing techniques for local contrast enhancement are designed to produce maximal contrast across the whole image. However, natural photos require selective improvements to emphasize the most valuable objects in the scene only. In this paper, we propose a local contrast enhancement algorithm that utilizes both local contrast distribution and saliency as an input. The proposed approach performs edge-aware processing that guarantees that the distinct objects get the correct contrast. We construct the bilateral grid to measure the contrast distribution in edge-preserving manner and obtain unique tone reproduction curves (TRC) of downsampled image pixels. We further tune TRCs using the saliency map in order to provide visually aesthetic contrast improvement. Finally, we upscale the contrasted image with slicing technique. The proposed approach can be utilized for automated aesthetic enhancement of images or emphasizing objects of interest.