Histogram Learning in Image Contrast Enhancement

Bin Xiao, Yunqiu Xu, Han Tang, Xiuli Bi, Weisheng Li; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 0-0

Abstract


In this paper, we propose a novel contrast enhancement method utilizing a fully convolutional network (FCN) to learn the weighted histograms from input images. In this method, the enhanced image references are not required. The training images are synthesized by randomly adding illumination on different regions in the source images to simulate the input images with poor contrast in different regions, and to enlarge the scale of training image set. And with this data-driven strategy for learning the underlying ill-posed illumination information of each pixel, a novel weighted image histogram is developed. It not only describes the distribution of pixel intensity, but also contains the illumination information of input images. Consequently, the proposed method can fast and efficiently enhance the regions with poor contrast and have the regions with acceptable contrast preserved, which keeps vivid color and rich details of the enhanced images. Experimental results demonstrate the effectiveness of our proposed method in comparison with some state-of-the-art methods.

Related Material


[pdf]
[bibtex]
@InProceedings{Xiao_2019_CVPR_Workshops,
author = {Xiao, Bin and Xu, Yunqiu and Tang, Han and Bi, Xiuli and Li, Weisheng},
title = {Histogram Learning in Image Contrast Enhancement},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}