Two-dimensional image processing without transpose

This paper provides a new solution to the cache efficiency problem in processing large two-dimensional image at both row and column directions, by implementing a special SDRAM controller in FPGA which can access image data at both directions efficiently and by employing some accessing rules in software which can guarantee cache coherence;. By contrast with traditional Corner Turn (transpose of matrix) method. Our design consumes less hardware resources by making additional Corner Turn Memory unnecessary, and has the processing time shortened by skipping the Corner Turn phase.