An FPGA implementation of real-time K-means clustering for color images

摘要: K-means clustering is a very popular technique, which used in numerous applications. In the k-means algorithm, each point dataset assigned to nearest cluster by calculating distances from centers. The computation of these time-consuming task, particularly for large and number clusters. order achieve high performance, we need reduce distance calculations efficiently. this paper, describe an FPGA implementation color images based on filtering algorithm. our implementation, when point, clusters are apparently not closer than other filtered out using kd-trees dynamically generated iteration clustering. performance system 512 × 640 480 pixel (24-bit full RGB) more 30 fps, 20–30 fps 756 average dividing 256