AI application in video: spiral optimizer based fast intra mode selection in HEVC

A codec is the most important element of any streaming application like Netflix or YouTube, but its complexity is increasing with its advancement. The codec, High Efficiency Video Coding (HEVC), selects the intra mode out of a total of 35 intra modes using the brute-force method. Therefore, the Spiral Optimization Algorithm (SOA) is utilized in this article to overcome the aforementioned limitation of HEVC. Firstly, the elements of SOA are efficiently mapped to the elements of HEVC. Secondly, the current best intra mode becomes the center of the spiral of SOA. Thirdly, the current best intra mode is considered optimal if the spiral completes three spiral-loops and the current intra mode still remains the best. Otherwise, the new best intra modes becomes the center of the spiral. Results show that SOA based fast intra-mode selection reduces the time complexity of HEVC by 34.09% while BD-BR (Bjontegaard Delta - Bit Rate) overhead is only 1.56%.