Application of high-performance normalized cross-correlation algorithm to image registration
Qingwei Zheng, Ziquan Zhong, Jun Zhao · International Journal of Applied Mathematics & Statistics/International journal of applied mathematics and statistics · 2014
The method of Normalized Cross-Correlation (NCC) is mostly adopted by image registration for scene matching systems. This theory method has been widely used; however, it has obvious drawbacks. It is necessary to shift or warp the relative images and it is also necessary to search where the most similar position is, and therefore the great complex computations will be produced during the period of searching and calculating the peak values pixel-by-pixel., especially under the condition of the source images with the high definition. In this paper, we propose a parallel technique for the high-performance applications based on FPGA+DSP architecture which supports the accurate registration of two images with high sampling factors. This design is established and implemented on the integral image issue and the Fast Ferial Transform theory. By applying the method, it will improve the computing performance and accomplish operations at the mirror cost rapidly and correctly. It is more applicable for the images with larger size.