Flash shadow detection using flash image pairs

When photographers taking pictures in low-light environments, they normally use flash light to enhance the illuminate. However, the use of flash may introduce unwanted artifacts, such as red-eyes and flash shadows. Detection and removal flash shadows from flash images can significantly improve the performance of several vision tasks such as object detection and recognition. A practical algorithm is proposed to detection flash shadows by using flash image pairs. The key hypothesis is that flash shadow edges appear only in flash images and the intensity values of flash shadow regions are much lower than corresponding regions in no-flash images. The developed algorithm consists of a three-tier process including pre-processing, flash shadow edge detection, and flash shadow detection. Experimental results demonstrate the effectiveness of the presented method as compared to previous methods.