Reduction of Glare in Images with Saturated Pixels

In this paper, we propose a deconvolution-based glare reduction method for HDR images. Specifically, we consider glared images containing saturated pixels and estimate the true radiance of these pixels followed by the deconvolution with the known glare spread function (GSF) to recover a glare-free image. The proposed method combines the image glare model with a dark channel prior for an accurate estimation of true radiance. The GSF is estimated separately in an offline calibration step. The qualitative and quantitative analysis of results on a variety of real and simulated glare images demonstrate the effectiveness of the proposed method.