Research and Realization of a Fast K-LT Algorithm
In order to overcome the large computing and the slow speed disadvantage for Karhunen-Loeve transform, this paper put forward a fast algorithm of Karhunen-Loeve transform. Firstly, the image matrix are cut equidistantly in columns direction by the choosing fit class interval to get sub-images of many same resolution; then these sub-images are stacked vertically. Secondly, the information relevance between all the sub-images is eliminated by the use of the statistical properties of the K-LT. Since this method recombined the information of image matrix and significantly reduced the dimension of feature space, so it makes a computational complexity decreases and a rapid increase in speed. For the image matrix which resolution is 1024x1024, the experimental results show that the FK-LT compared to the conventional KLT, which the operation time can be shortened about 50 times, it can meet the real time needs.
Karhunen-Loeve transform Fast Karhunen-Loeve transform Cutting Stacking
Wan-hong Niu Jing Zao Guo-jun LIU Shao-hua Yang
School of Remote Education, Ningxia University, Yin chuan, China School of Mathematics and Computer, Ningxia University, Yin chuan, China
国际会议
哈尔滨
英文
350-356
2012-05-19（万方平台首次上网日期，不代表论文的发表时间）