A Review:SIFT Hardware Implementation For Real Time Feature Extraction

In pattern recognition and image processing, feature extraction is simple form of dimensionality reduction. The transformation of the input data's into a set of features is known as feature extraction. The large set of data's is to be analyzed and performed accurately from the features of the input data. By using SIFT (Scale Invariant feature transform) the hardware resources can be minimized and it could be performed as a process of parallel and the pipeline based VLSI architecture. We propose two parallel SIFT feature extraction algorithms using general multi-core processors, as well as some techniques to optimize the performance on multi-core. The proposed architecture led to a 6.7x faster speed on a dual-socket, quad-core system, which facilitated an average 45 frames/second for a VGA (640×480) video. Some implementations and accelerations of SIFT feature extraction on graphics processing units (GPUs) were introduced. With the parallelism and powerful computational ability, they achieved high processing speed.