面向互联网应用的图像LBP算法GPU并行加速.pdfVIP

面向互联网应用的图像LBP 算法GPU 并行加速 1+ 1 1 1 2 王香荣 , 高 飞 , 李 钦 , 刘 轶 , 张 阔 1(北京航空航天大学 中德联合软件研究所,北京 100191) 2(北京搜狗科技发展有限公司,北京) GPU-based Acceleration of Local Binary Pattern Algorithm for Internet Applications * 1+ 1 1 1 2 WANG Xiang-Rong , GAO Fei , LI Qin , LIU Yi , ZHANG Kuo 1(Sino-German Joint Software Institute, Beihang University, Beijing 100191, China) 2(Beijing Sogou Co., Ltd, Beijing, China) Abstract : Many Internet applications need to process massive images. As a typical image texture feature extraction algorithm, Local Binary Pattern (LBP) is widely used in image retrieval and other fields. This algorithm is relatively complex, so it faces the performance challenges on processing massive images on the Internet. This paper presents parallelization implementation of extracting Local Binary Pattern texture feature using the Graphics Processing Unit (GPU), especially an accelerating scheme for the processing of massive images. The GPU can compute the LBP feature of multiple images in parallel. And asynchronous transmission is used to perform the data replication and kernel function of multiple images in parallel. The paper tests the parallel processing of single image and multiple images using the GPU, and compares the performance of the program on the GPU with the one on the CPU. It is shown that the parallel processing for different resolution multiple images can achieve 58 times speedup. Key words: GPU; Local Binary Pattern; asynchronous transmission; Internet 摘 要: 很多互联网应用需要进行海量的图片处理。LBP(Local Binary Pattern)算法是一种图像纹理特征提取算 法，广泛用于图像检索等领域，但该算法较为复杂，在处理互联网环境下的海量图片时面临着性能挑战。本 文采用GPU(Graphic Processing Unit)对LBP 算法进行了并行加速，特别针对海量图片处理设计了加速方案， 使 GPU 同时进行多幅图像LBP 特征的并行提取，并采用异步传输方式