Intraprediction Complexity Control Algorithm Based on Visual Saliency

Intraprediction is one of the most complex parts High-Efficiency Video Coding (HEVC), because it selects best prediction mode by calculating cost every Unit (CU), which provides higher complexity intracoding. Visual saliency map can show attention regions human eyes, generated certain static and space-time detection method. By analyzing percentage coding time for different size CU, relation visual CU depth, an intraprediction control algorithm based on proposed in this paper. Based feature video target level, threshold adapted to determine whether current processing should be split into smaller CUs or division stopped early. Three samples were compared other algorithm, has better performance PSNR, BitRate, time. Experimental results that effectively with minimal loss applied a number scenarios, such as real-time coding.