摘要
彩色局部角度二值模式(CLABP)可以有效地提取彩色图像中的纹理特征,但是算法复杂、计算量大。针对这一问题,采用GPU实现CLABP特征提取和表示的并行方法。该方法一方面使用异步处理的方式实现CLABP的并行加速,另一方面采用共享内存的形式减少读取数据的次数。为了验证该方法的有效性,在Outex纹理图像数据库上与CPU程序性能进行对比,结果表明,GPU实现方法可以提升加速比约25倍。
Color local angel binary patterns （CLABP）method can extract color texture feature effectively. But the algorithm is complex and has large amount of calculation. In order to solve the problem, this paper proposes a GPU implement for color local angel binary patterns texture feature extraction and presentation. On one hand, this method use asynchronous processing way to realize CLABP parallel acceleration. On the other hand, it also improves the implement via shared memory to reduce the number of reading data. In order to verify the validity of proposed method, we compare our method with CPU implement on Outex texture dataset. The experiment results demonstrate that the acceleration rate can achieve 25 times.
出处
《微型机与应用》
2014年第16期37-39,共3页
Microcomputer & Its Applications
关键词
GPU
彩色局部角度二值模式
异步处理
GPU
color local angel binary pattern（CLABP）
asynchronous processing
