Color Local Angel Binary Patterns Method Based on GPU Implementation

Color local Angel binary patterns method can extract color texture feature effectively. But the algorithm is complex and large amount of calculation. In order to solve the problem, this paper proposed a GPU implement for Color Local Angel Binary Patterns texture feature extraction and presentation. On one hand, our method use asynchronous processing way to realize CLABP parallel acceleration. On the other hand, we also improve 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 experimental results demonstrate the acceleration rate can achieve 3 times.