Parallel processing for accelerated Mean Shift algorithm based on TBB

Image segmentation as a main applying field in parallel computing with high performance, its time complexity and real-time requirements of algorithm needs to continue to improve computer hardware technology and parallel computing algorithm. Mean Shift algorithm is relatively classical in image segmentation fields, which needs no prior knowledge in the process and is an unsupervised segmentation process, attracting widespread attention for its good applicability. The paper makes a parallel improvement of Mean Shift algorithm using TBB on multi-core. The paper first analyzes the most time-consuming part Mean Shift clustering in the process of Mean Shift image segmentation, then makes a parallel improvement of Mean Shift clustering base on TBB and gets a preferable accelerating effect.