基于位图掩码的GPU图像合成优化  

Bitmap-based Multi-GPU Image Compositing Optimization

作　　者：刘华海[1] 王攀[1] 李思昆[1] 蔡勋[1] 曾亮[1] 王文珂[1] 

机构地区：[1]国防科学技术大学计算机学院,长沙410073

出　　处：《系统仿真学报》2012年第9期1940-1944,共5页Journal of System Simulation

基　　金：国家"973"项目(2009CB723803);国家自然科学基金项目(61170157)

摘　　要：重点对多GPU系统上图像合成优化方法进行研究,提出一种基于压缩掩码位图的图像合成优化方法。该方法以二进制位表征像素的有效性状态,实现了图像有效像素的编码压缩,并在压缩编码基础上定义了4种位运算操作,能够快速完成图像重叠区域的判别,实现了多GPU系统上高分辨率图像的快速合成。实验结果表明,与传统基于包围盒及RLE编码的图像合成优化方法相比,压缩位图编码方法能够有效提高图像合成速度近40%左右。Image compositing optimization was studied, and a novel compressed bitmap-based image compositing method was proposed. The state of pixels was represented by binary bit, and an encoding approach was implemented to compress active pixels in an input image. Four bit operators were defined to conveniently acquire the overlap region of two input images, and then efficient composition of high-resolution images could be implemented by the methods. The experiments shows that the bitmap-based compositing can achieve about 40% faster than previous solutions such as bounding box and RLE.

关 键 词：并行绘制 压缩掩码位图 多GPU 图像合成 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]