Line Buffer Reduction for LUT-Based Real-time Image Inverse Warping

Abstract

Inverse warping is widely adopted to rectify the image with non-linear distortion, which is caused by the perspective feature of optical lenes. This scheme stores warping coordinate into a look-up table (LUT) to avoid complex coordinates calculations, but requires large amounts of line buffers to place image data to perform inverse warping. To reduce the line buffer usage, this paper presents a line buffer algorithm based on reading extended, which utilizes the horizontal banking to extend the reading period. The algorithm guarantees the write and read pointers dynamical jumping to provide the space for up-mapping and down-mapping. The dedicated control scheme can reduce nearly half of the memory usage in contrast to a general circular buffer of maximum offset. The image rectification module has been implemented on a FPGA to remove lens distortions in real time. The experimental results prove that the proposed line buffer algorithm is highly effective in both memory reduction and distortion rectification.