Low Memory Image Reconstruction Algorithm from RAW Images

摘要

In this paper, we present a fast and low memory image reconstruction algorithm from a burst of RAW images. Existing algorithms produce high-quality images but the number of input images is limited by severe computational and memory costs. Our algorithm processes the images sequentially so that the memory cost only depends on the size of the output image. Data are combined using classical kernel regression and blur is removed by applying the inverse of the corresponding asymptotic equivalent filter that we introduce. In addition, we propose an accurate and efficient registration method for mosaicked images. We verify the performance of our algorithm on synthetic and real images. For a large amount of data, the results are similar to slower memory greedy methods.