Automatic local exposure correction using bright channel prior for under-exposed images

• Bright channel prior—a new prior for accurate estimation of the local exposure.
• A local exposure correction method based on the bright channel prior.
• A spatial-variant denoising algorithm to reduce the noise in exposure correction.

We address the problem of exposure correction for under-exposed images in this paper. We propose the bright channel prior based on the statistics of well-exposed images. Using the proposed prior, we are able to estimate the relative exposure in local image regions. A well-exposed noise-free image is then recovered after exposure correction followed by non-uniform denoising and detail enhancement. Our experiments on under-exposed images of various scenes demonstrate the effectiveness of the bright channel prior. Our user study shows that the results generated by our exposure correction method are preferred over existing methods.