Parallel Implementation of Good Feature Extraction for Tracking on the Cell Processor with OpenCV Interface

One of the most widely used schemes to extract feature points suitable for tracking in computer vision is "good features to track\´\´. In this paper, we propose parallel implementation of the good feature extraction scheme optimized for the cell processor, which is one of the latest high performance embedded processors. By utilizing the computational power of cell suitable for image processing, we achieve high-speed computation of the operation. Experimental results show that our implementation with 6 SPEs can compute the feature point extraction in 1.4 ms when the input image size is 640 times 480 pixels. This is about 5 times faster than the computation on Intel(R) Core(TM)2 Duo CPU E6850 @ 3.00 GHz with Intel Integrated Performance Primitives. This work is part of the CVCell project which is a software library compatible with OpenCV library optimized for the Cell processor.