A Fusion Algorithm for Flash and No-Flash Images under a Low-Light Environment

Pictures which are taken under a dim light condition might be noisy or relatively dark. Generally, we open the flash to increase the exposure to obtain a flash image. It has high definition but often changes tone of the image. On the contrary, a no-flash image remains the tone but has low definition. Therefore, it is natural to explore a fusion algorithm to combine the advantages of the two kinds of images in order to improve the final image quality. A common fusion algorithm is composed of matching and filtering. However, the conventional matching algorithms are only suitable for the same illumination situations. There may be few feature points while matching based on flash and no-flash images of the same scene. After matching, we employ the filter based on multi-scale detail extraction to obtain the details and it may lead to color distortion. Despite such problems, the conventional matching and filtering algorithms have their own advantages. In order to inherit their advantages on matching and filtering and address the potential problems, we propose an advanced matching algorithm ASIFT and an advanced filtering algorithm AWLS in this paper. Experimental results demonstrate that ASIFT can increase the number of matching pairs and ALWS can eliminate color distortion and improve the details.