Fast CTU Splitting Using Edge Detection in Video Coding

The HEVC/H.265 (High Efficiency Video Coding) standard offers notable bitrate improvements over the preceding AVC/H.264 (Advanced Video Coding) standard, but it also introduces exceedingly complex encoding. A major factor contributing to this complexity is the partition of coding tree units (CTUs), which involves a brute-force search of numerous combinations for rate-distortion optimization (RDO). The block partitioning strategy Quad-Tree (QT), to achieve a balance between bit rate and quality for each coding unit (CU) of each frame within the video stream, is principally responsible for the complexity overhead. To tackle this issue, we propose a technique to reduce the time complexity of intra-mode HEVC. Instead of relying on RDO, our approach utilizes an edge detection model to detect crucial edges within a frame. Edges having sensitive information are partitioned to the maximum CTU depth. In this way, we provide early CTU partition decisions. Experimental results demonstrate that our approach reduces intra-mode HEVC encoding time, over 12 standard test video sequences, up to 48.04% and on average 41.83% with Bjøntegaard delta bit-rate of 5.24% using Canny edge detection method. Similarly, we achieve complexity reduction time up to 72.98% and on average 58.41% at the expense of 3.93% Bjøntegaard delta bit-rate using Sobel Edge detection method. This proves that the proposed approach is more time-efficient than other state-of-the-art methods.