基於主觀視覺之H.264可調式視訊編碼空間層位元率分配機制

摘要

H.264可調式視訊編碼是最新一代之可調適視訊編碼標準。其提供三種可調適架構包含時間可調性，空間可調性與品質可調性，同時滿足不同應用服務之需求。其中空間層可調性架構因其可同時提供不同解析度之視訊資料供使用者選擇而最廣泛的被應用。而在實際的應用層面上，如何有效率的分配位元率給不同解析度的空間層，是十分重要的議題。

現行的SVC編碼架構利用低解析度空間層資訊對高解析度層做編碼，因此低解析度空間層信號被視為最重要，因此編碼以維持最低解析度空間層的品質，即固定低解析度空間層量化參數繼而調整高解析度空間層畫面量化參數之方式來達到頻寬限制的要求。然而，根據實驗結果顯示人眼視覺系統對於高解析度畫面有較高的敏感度，因此位元率下降所造成的畫面失真影響在高解析度空間層，將更為嚴重。在頻寬共享的環境中，對不同解析度的使用者而言，應要能享有幾近相同的視訊服務品質。本論文即以在頻寬有限的條件下，分配位元率給不同解析度空間層，提供各個不同解析度需求使用者最佳之視訊品質。

本論文提出一套在網路資源有限的情況下，考量使用者主觀視覺感受的H.264可調式視訊編碼空間層位元率分配機制，改善傳統固定各層QP的位元率分配機制下，高解析視訊擁有較低的主觀視覺品質。我們降低人眼視覺中較不靈敏的低解析度視訊品質，並提升高解析度視訊品質，期望不同解析度的使用者（或當使用者變動使用距離時），能擁有接近的主觀視覺品質，實驗結果顯示，本論文所提的位元分配機制可有效率的分配位元率使各解析度空間層達到相近的主觀視覺品質，且對於不同視訊資料在不同的頻寬限制下均表現良好，更進一步提升了以主觀視覺品質為目標之下的頻寬使用率。

　

H.264/SVC Rate Allocation for Spatial Scalability Based on Perceptual Quality Metric

Abstract

H.264 scalable extension (SVC), which is constructed based on H.264/AVC, is the most recent scalable video coding standard. It offers three scalabilities in spatial, temporal, and quality, to meet multiple requirements simultaneously. Spatial scalability that can support multiple display resolutions with a wide range of bitrates is used widely. How to efficiently allocate a given total bitrate among multiple layers under the bandwidth constraint is an important issue and should be solved at first.

The base layer is usually treated more important than the enhancement layer because the information in base layer will often be re-used in enhancement layers. Therefore, under a bandwidth constraint, we usually run SVC by fixing the Qutization Parameter (QP) or the bitrate of the base layer while adaptively adjusting the ones of the enhancement layers. However, it is observed that Human Visual System (HVS) is more sensitive to higher resolution videos; in other words, the quality degradation at higher layers to human eyes would be more serious than that at lower layers. The main objective of this work is to achieve best and equal quality for each resolution layer under a given bandwidth constraint.

This thesis proposes a rate allocation method for SVC spatial scalability based on perceptual quality metric. We utilize the subjective metric, instead of conventional objective measurement PSNR, to measure video quality. Each resolution layer is measured by the quality metric and allocated with the corresponding rate to have similar quality. The disadvantage of the conventional fixed QP scheme that the higher resolution layer has worse subjective quality is improved. In simulations, several video sequences with various total rate constraints are experimented. The proposed method can efficiently allocate the rate for each layer with almost the same video quality in subjective measurement..