基於HEVC畫面內編碼特徵副本偵測

摘 要

當今多媒體技術蓬勃發展，網路及行動裝置的普及也造就海量資料的暴增這也意味著龐大的數位資料量將面臨儲存與管理上的問題， 因此多媒體資料的管理、應用及相關議題也逐漸成為備受重視的議題。因此副本偵測（copy detection）就成為一個重要的課題，則副本偵測的相關研究繁多，而壓縮域中的研究也不在少數但仍有缺失，特別是將影像全部解壓縮後再從影像擷取特徵，會耗費相當多地運算時間與儲存空間。

本論文針對內容檢索，利用HEVC解碼器於壓縮域上的特徵進行抽取，在進行特徵比對前將畫面內預測模式（intra prediction mode）和殘餘值（residual coefficients）精化作為本研究之比對特徵以提高比對的效能。實驗結果顯示本研究中所提出之方法在評分標準Precision, Recall, Accuracy, F1 皆可達到與像素域相同的檢索成效。

Copy Detection Utilizing HEVC Intra Coding Features

ABSTRACT

The rapid growth of internet and online video data bring the difficulties not only to database management but also to digital right management. In the past decade, numerous algorithms were developed for video copy detection.

Because almost all digital video data exist in a compression format, many copy detection techniques operated in compression domain were proposed. However, these methods did not perform well with the sequences that had different quantization parameter (QP).

This thesis focuses on the content-based copy detection in compression domain, especially with the capability resisting multi-QP and frame sizes. The extracted features are based on the I-frame coding information in HEVC that is the newer coding standard. We propose to employ the direction mode and residual coefficients as the texture feature.

The experimental results show that the proposed method can reduce the resource consumption and archive the similar performance with the pixel domain approach.