Performance Evaluation of Feature Extraction Algorithm on GPGPU

Nvidia's GPGPU based Compute Unified Device Architecture (CUDA) is a software platform for massively parallel high-performance computing on GPU. It provide several key abstractions- a hierarchy of thread block, shared memory, and barrier synchronization. This model has proven quite successful at programming multithreaded many core GPUs and scale transparently to hundreds of cores: many industry and academia are already using CUDA to achieve speedups on production and research codes. This paper analyze distinct feature of CUDA GPU, summarizes general programming mode of CUDA. This paper presents image processing algorithm i.e. Canny Edge detector which is used as pre-processing steps in many computer vision application as an optimal edge detection algorithm. Detailed comparison of parallel and sequential algorithm implementations is also presented.