Abstract
In order to design a real-time embedded system for catadioptric omni-directional images, a look-up table approach for omni-directional images was used on FPGA. Combined with FPGA features, some methods were used to improve the speed of panorama unrolling, including dividing image into blocks and unrolling the image blocks, time-hidden data access technology and FPGA pipeline technology, etc. Experiments indicate that the methods make the unrolling achieve a level of 100 frames per second, when dealing with 1024×768 YUV 4∶2∶2 source image to 1 280×256 YUV 4∶2∶2 target image. It is 12 times faster than a look-up table panorama unrolling based on single pixel.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have