Title :
Real-time Object Segmentation based on GPU
Author :
Lee, Sun-Ju ; Jeong, Chang-Sung
Author_Institution :
Sch. of Electr. & Comput. Eng., Korea Univ., Seoul
Abstract :
In this paper we present a new novel GPU-based algorithm for real time object segmentation by employing several techniques which make use of framebuffer, swizzle operator, stencil buffer and the other factors on GPU. Frame buffer is used for the fast processing of image on GPU without going back and forth to the CPU, swizzle operator for the efficient and fast execution of vector operations for object segmentation, and stencil buffer for the fast computation of the masked area for the detected object, thus speeding up the whole algorithm sharply. Moreover, the computation of the histograms and bounding boxes make our algorithm very simple and fast by using ´gather´ operation uniquely supported by GPU. Our experimental results have shown that our algorithm is much faster than CPU based one
Keywords :
image segmentation; microprocessor chips; object detection; real-time systems; GPU-based algorithm; framebuffer; real-time object segmentation; stencil buffer; swizzle operator; Brain modeling; Computer graphics; Computer vision; Histograms; Image processing; Layout; Object detection; Object segmentation; Real time systems; Surveillance;
Conference_Titel :
Computational Intelligence and Security, 2006 International Conference on
Conference_Location :
Guangzhou
Print_ISBN :
1-4244-0605-6
Electronic_ISBN :
1-4244-0605-6
DOI :
10.1109/ICCIAS.2006.294232