Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research and Implementation of GPU-based visual image processing method

Author HeXiaoMan
Tutor ZhaoZhengWen
School Southwest Petroleum University
Course Applied Computer Technology
Keywords GPU CUDA Image Processing Parallel computing
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 210
Quotes 1
Download Dissertation

Compute speed is always a key factor constraining the use of graphics in digital and image processing, With the rapid development of graphics technology, the complexity of image processing and computing capacity make a progressive success. However, current image processing speed based on CPU is relatively slow. How to obtain a high-speed computing by current hardware is a new topic.This subject root in 2008-2010 National Science and Technology Program " large oil and gas fields and coal bed methane development " subordinate project " offshore oilfield effective development new technology " (Project number:2008ZX05024) subordinate project " Multibranched diversion Balanced take out gravel technology " (Project number 2008ZX05024-03) subordinate project number " Comprehensive oil store research of multibranched diversion Balanced take out gravel technology" (subordinate Project number: 2008ZX05024-03-01-03).This subject mainly research the image processing for the program-availability of loosen gritstone microcosmic drive displace granule move and seepage visualization image analysis software.This article first discuss the way by using OPENGL matrix operations, cluster systems, and multi-processing system to improve the graphics processing speed and set up a corresponding test environment and test program and also the result that OPENGL matrix operations, cluster systems and multi-processing system can not process image immediately.Based on theoretical analysis, the article discuss the algorithms by CUDA programming sharpening, edge recognition and image segmentation algorithm to speed up image processing.Then this article detail and demonstrate the availability of CUDA, including resolution of memory access conflict, Division of the CUDA internal module and thread and rational use of GPU texture memory. Through these detail studies further improve the way of speed up image processing.Finally, we get conclusion, system speed up by using this method in " loosen gritstone microcosmic drive displace granule move and seepage visualization image analysis software. " and it works well.

Related Dissertations
More Dissertations