基於空間特徵與摺積神經網路於 H.266/FVC 視訊畫面內編碼快速分割決策

摘要

跟隨著網路和多媒體的科技迅速成長，我們人類平日生活的高解析度視訊重要性一日千里，最近市面已出現許多4K高解析度的視訊，可以看見未來高解析度視頻勢必會成為主流，而目前的最新視頻壓縮標準H.265/HEVC已經逐漸不夠能用，因此ITU-T VCEG和ISO/IEC MPEG共同組成的JVET (Joint Video Exploration Team)來制定新一代的視訊壓縮標準H.266/FVC (Future Video Coding)從2015年開始討論，且預計2021年正式發佈變國際視訊壓縮標準。

H.266/FVC相差於H.265/HEVC採用QT(QuadTree)改成QTBT(QuadTree plus Binary Tree)，不但將複雜的CU、PU和TU的組成元素除去，且能夠支援最大256x256到最小8x8的正方形區塊，以依據更多不同大小畫面的紋理特性編碼。QTBT的架構雖比QT提供更好的編碼效能，但其預測數量的增加造成在執行畫面編碼時間提高了5.6倍，於是針對畫面內編碼，如何發展增快CU編碼時間下的決策，這是非常重要的議題。

本論文結合近年十分熱門的人工智慧系統(Artificial Intelligence, AI)，提出基於摺積神經網路於H.266/FVC視訊編碼畫面內快速分區預測決策法。主要分為兩部分探討：首先在第一部份針對預測模型的訓練和訓練資料的選擇來討論；在第二部份則是將訓練好的預測模型結合至H.266/FVC壓縮參考軟體當中來執行編碼。

關鍵字 : 未來視訊壓縮編碼、編碼單位、畫面內編碼、分區預測、摺積神經網路、空間特徵

　

Fast Partition Decision for H.266/FVC Video Intra Coding Based on Spatial Features and CNNs

Abstract

With the rapid growth of Internet and multimedia technology, the high-resolution video of our daily lives is becoming more and more important. Recently, there have been many 4K high-resolution videos on the market. It can be seen that high-resolution video will become the mainstream in the future. However, the current latest video compression standard H.265/HEVC has gradually become insufficient. Therefore, the JVET (Joint Video Exploration Team) formed by ITU-T VCEG and ISO/IEC MPEG has developed a new generation of video compression standard H.266/ FVC (Future Video Coding) has been discussed since 2015 and is expected to be officially released in 2021 to become an international video compression standard.

H.266/FVC is different from H.265/HEVC. QT (QuadTree) is changed to QTBT (QuadTree plus Binary Tree), which not only removes the complex elements of CU, PU, and TU, but also supports the largest 256x256 to the smallest 8x8 Square blocks are coded according to the texture characteristics of more pictures of different sizes. Although the QTBT architecture provides better coding performance than QT, the increase in the number of predictions has resulted in a 5.6 times increase in the execution picture coding time. Therefore, for intra-picture coding, how to develop decision-making under CU coding time is very important.

This thesis combines the popular Artificial Intelligence (AI) system in recent years, and proposes a fast partition prediction and decision-making method based on the convolutional neural network in the H.266/FVC video coding frame. The discussion is mainly divided into two parts: First, the training of the prediction model and the selection of training data are discussed in the first part; the second part is to integrate the trained prediction model into the H.266/FVC compression reference software to perform coding.

Keyword: Future Video Coding (FVC), Coding Unit (CU), Intra Coding, Partition Prediction, Convolutional Neural Network (CNN), Spatial Feature