A new method of flash image shadow detection and patch-based correction of shadow regions

While capturing an image under flash light, we obtain a relatively sharp and noise-free image but the presence of sharp shadows at silhouettes is unwanted and sometimes it also reduces the overall visual appeal of the image. In this paper, we present a shadow detection as well as removal algorithm to detect and remove shadows at silhouettes from images captured under flash light. We use a pair of flash and no flash images for performing shadow detection and recover the shadow areas to produce an enhanced flash image. This problem has not been addressed explicitly before in literature with major emphasis on flash image enhancement. We compare the accuracy of proposed shadow detection algorithm with an existing method in literature. It is shown that proposed method of shadow detection gives more accurate results and the proposed shadow recovery method gives subjectively better results compared to pixel-wise recovery method.

Paper

Full text

PDF

A new method of flash image shadow detection and patch-based correction of shadow regions

Semantic Scholar · Computer Science · 2015

Abstract

While capturing an image under flash light, we obtain a relatively sharp and noise-free image but the presence of sharp shadows at silhouettes is unwanted and sometimes it also reduces the overall visual appeal of the image. In this paper, we present a shadow detection as well as removal algorithm to detect and remove shadows at silhouettes from images captured under flash light. We use a pair of flash and no flash images for performing shadow detection and recover the shadow areas to produce an enhanced flash image. This problem has not been addressed explicitly before in literature with major emphasis on flash image enhancement. We compare the accuracy of proposed shadow detection algorithm with an existing method in literature. It is shown that proposed method of shadow detection gives more accurate results and the proposed shadow recovery method gives subjectively better results compared to pixel-wise recovery method.

Similar papers

© 2026 NYSGPT2525 LLC