Correction of Over-Exposed Images Captured by Cell-Phone Cameras

摘要

Due to the limited dynamic range of commonly used cell-phone camera sensors, the entire or part of an auto-exposure (AE) image may appear over-exposed. In this paper, we present an over-exposure correction algorithm which utilizes one low-exposure (LE) image in addition to the AE image. It consists of registration, fusion, consistency verification, and merging components. Registration is achieved by using a one-dimensional grid correlation approach. A linear weighted summation of the AE and LE images is utilized to achieve fusion. A mechanism based on image entropy and brightness is developed to appropriately fuse subimages or blocks. Consistency verification and Gaussian smoothing are then applied to remove blocking artifacts. The results obtained show that the developed algorithm is more effective than the existing and comparable overexposure correction algorithms.