Spatial Error Concealment using Log-Polar Domain Based Image Inpainting

利用基於對數極座標影像修補之空間域錯誤隱匿

吳瑋晟(Wei-Cheng Wu)

指導教授 ： 張隆紋

國立清華大學/電機資訊學院/資訊工程學系/碩士(2009年)

https://doi.org/10.6843/NTHU.2009.00621

摘要

Error concealment is one of the techniques that have been adopted to recover corrupted information for video transmission. In this article, a novel I-frame error concealment approach based on the concept of exemplar-based image inpainting is proposed. The proposed algorithm employs the log-polar transformation and least squares matching on the exemplar-based image inpainting. The log-polar transformation makes the patches in the exemplar-based image inpainting invariant to scale and rotation, while the least squares matching method refines the patch with affine transformation. In this way, the proposed inpainting algorithm is superior to deal with strong structured image. In order to meet the property of block-based coding in H.264 standard, the proposed inpainting algorithm is transferred to block-based manner. The experimental results show that the proposed algorithm outperforms the default solution in H.264 standard and conventional exemplar-based image inpainting in subjective quality.

關鍵字

H.264 ； 錯誤隱匿 ； 影像修補 ； 對數極座標轉換 ； 最小二乘匹配法

並列摘要

錯誤隱匿是一種常用於視訊傳輸時修補毀損資訊的技術。本文中，運用基於範例影像修補(exemplar-based image inpainting)的概念提出了一個新的I-frame 錯誤隱匿方法。提出的演算法將指數-極座標轉換(log-polar transformation) 及最小二乘匹配法 (least squares matching) 應用在基於範例影像修補當中。指數-極座標轉換將影像修補當中的補丁(patch)達到旋轉及尺度的不變性，而最小二乘匹配法利用仿射轉換(affine transformation)將補丁更進一步細緻化。利用這樣的方法，所提出的影像修補演算法能夠更優秀地處理具有強烈結構的影像。為了滿足H.264標準中的區塊式編碼特性，提出的影像修補演算法也被轉換成區塊式的修補方式。實驗結果顯示提出的影像修補演算法在視覺效果上優於H.264標準中的預設方法以及傳統的基於範例影像修補方法。

並列關鍵字

H.264 ； error concealment ； inpainting ； log-polar transformation ； least squares matching

參考文獻

[2] C. Chen, Y. Liu, Z. Yang, J. Bu, and X. Deng, “Multi-frame error concealment for H.264/AVC frames with complexity adaptation,” IEEE Transactions on Consumer Electronics, vol. 54, issue. 3, pp 1422-1429, August 2008.

[3] T.H. Wu, G.L. Wu, C.Y. Chen, and S.Y. Chien, “Enhanced temporal error concealment algorithm with edge-sensitive processing order, ” IEEE International Symposium on Circuits and Systems, pp. 3466-3469, May 2008.

[4] A. Sirikam and W. Kumwilaisak, “New Spatial Error Concealment using Dynamic Texture Estimation and Geometric Interpolation, ” IEEE International Conference on Multimedia and Expo, pp. 120-123, July 2007.

[6] Y. Shi, X. Zhu, J. Xia, and H. Yin, “A Fast and Efficient Spatial Error Concealment for Intra-coded Frames,” Congress on Image and Signal Processing, vol. 1, pp. 264-267, May 2008.

[7] Z. Wang, J. Ming, and B. Fan, “Fast Best Neighborhood Matching Algorithm for Intra Block Error Concealment in H.264/AVC,” Congress on Image and Signal Processing, vol. 1, pp. 559-563, May 2008.