具備多參考源高頻補償的視訊超解析技術

隨著科技進步，人們對影像畫質的要求也越來越高。不只是家用電視畫質從數年前的HD、FHD甚至到現今已不算少見的UHD (Ultra High-definition，4K) 。如此快速地提升，連帶行動錄影裝置 (如：手機、行車紀錄器) 和定點監視裝置(大賣場或十字路口的監視器等) 錄影品質也增進許多；相對地，需要儲存和計算的影片資料量也日益龐大，但是儲存設備的容量卻不一定跟得上資料量的成長速度。本論文於編碼端將影片降取樣，降取樣法採用小波轉換後的LL頻帶。降低影片解析度以減少傳輸需要的頻寬和儲存所需的空間，並保留部分高解析畫面作為關鍵畫面，以混合型影片進行傳輸。在解碼端實現多重參考畫面 (包含前後關鍵畫面與已經完成超解析的畫面) 重建高解析影片，如此亦可以降低對傳輸頻寬和儲存空間的需求。本論文主要先探討前後關鍵畫面和非關鍵畫面之低頻影像間的時間相關性，依相關程度建立權重式多參考源的高頻補償，模擬的結果與只參考前後關鍵畫面的權重式高頻補償相比，至少有1.4dB以上的增益。

關鍵字

超解析；結構相似性；前項超解析結果；權重式多參考源；關鍵畫面

並列摘要

The resolutions of video camera and display are improved dramatically because of the rapid innovation of Digital Multimedia Technology in recent years. This improvement can benefit many applications such as medical imaging, satellite imaging and digital surveillance systems. Although, high-resolution videos are very popular in the daily life, the digital video processes about sampling and compression are still resource-consumption, especially for storage, network and energy. Super-Resolution technology, which can reconstruct the high-resolution video from the low resolution set (the down-sampled video), could avoid the high-cost issues of camera and storage. In this paper, the encoder (a dual video recording system) can generate a non-mixed compressed data (sparse high-resolution and normal low-resolution produced by LL-band from Discrete Wavelet Transform). At decoder, considering the key (forward and backward) and the last SR reconstructed frames, more similar blocks are found to realize the multiple-frame motion estimation by analyzing the correlation. Therefore, the weighted motion compensation from multiple frames is realized to reconstruct the high-resolution sequence. Experimental results show that the gain of quality is more than 1.4dB compared with the weighted motion compensation with only key frames.

並列關鍵字

Super-Resolution ； Structural Similarity ； last SR result ； Multi-frames ； Key frames

參考文獻

[22] 柯敬奎，基於部分高解析關鍵畫面傳輸之視訊超解析重建技術，國立中正大學電機工程學系碩士論文，中華民國一百零一年七月

[2] Qiang Zhou, Li Song and Wenjun Zhang, “Video Coding With Key Frames Guided Super-Resolution,” Proc. Of Pacific-Rim Conf. on Multimedia (PCM-2010), Shanghai, China, Sept. 2010.

[15] Yao Wang, Jôrn Ostermann, and Ya-Qin Zhang , “Video Processing and Communications”, Pearson Education Taiwan Ltd, Jan. 2008.

[5] Nabil G. Sadaka and Lina J. Karam, “Efficient Super-Resolution Driven By Saliency Selectivity,” IEEE 18th International Conference on Image Processing (ICIP), pp.1197 – 1200, Sept. 2011.

[6] K. Takahashi, T. Naemura, M. Tanaka, “Rate-distortion analysis of super-resolution image/video decoding,” IEEE International Conference on Image Processing (ICIP), pp. 1629 – 1632, Sept. 2011.

國際替代計量

具備多參考源高頻補償的視訊超解析技術

未授權

主題瀏覽