MV和3D-HEVC編碼的新應用－多重亮度紅外線影片之編碼

Abstract

夜間監視系統在現今已經被廣泛的應用在各種場所，而多亮度紅外線打光器也已經被提出用來解決單一亮度下，遠距離過暗而近距離過曝的問題。但是，多亮度的特性會造成低落的影像編碼效率。 本篇論文中，將針對裝設多亮度紅外線打光器之攝影機，所拍攝下來的影片嘗試進行更有效率的影像編碼。我們使用了HEVC (High Efficiency Video Coding) 這個新一代的影像編碼標準，作為主要的編碼工具。首先，將原影片直接用Single View的方式送入HEVC編碼，並作為效率的基準點。接著將影片根據打光器之不同channel，分割成數段亮度固定的影片，並當作不同的View，使用HEVC的Multi-view以及3D模式進行編碼。由於分割後的固定亮度影片，很適合Multi-view及3D模式的幀間預測行為，故最後我們能夠得到較佳的編碼效率。另外，我們將討論在不同場景下，如何改善編碼效率及因應對策。

Nighttime surveillance has been widely used in several places nowadays, and Multi-Intensity Infrared Illuminator has been proposed to solve the under-exposure and over-exposure problems at long and short distance between the camera and target objects, which is caused by fixed brightness. However, the character of multi-intensity will result in low efficiency in video coding. In this thesis, we will try to perform video coding in a more efficient way toward the videos recorded by camera equipped with Multi-Intensity Infrared Illuminator. HEVC (High Efficiency Video Coding), which is the new generation of video coding standard, is adopted for it for its outstanding compression performance. First, we directly send the original video as a single view into HEVC to do video coding, and set the coding performance as baseline. Next we separate the original video into some brightness-fixed videos according to the different channel of Multi-Intensity Infrared Illuminator, and send them into Multi-view and 3D mode of HEVC to do the video coding. Because the video after separating are suitable for the inter-view prediction of Multi-view and 3D mode of HEVC, finally we will get better performance compared to single view case. Besides, we will discuss the strategies and how to improve the coding efficiency and in several different scenes.