植基於小波轉換之動態影像尺寸調整技術之研究

Content-Aware Image Retargeting Based on Wavelet Decompose

施存興(Tsun-Hsing Shih)

指導教授 ： 許永洲

國立臺中科技大學/設計學群/多媒體設計系碩士班/碩士(2010年)

https://doi.org/10.6826/NUTC.2010.00055

摘要

基於影像內容(Content-Aware)的影像尺寸調整技術(Image Retargeting)，有別於傳統方法，基於影像的內容計算其重要性的區塊(Region-of-Interest, ROI)，能在影像長寬比率的改變後，將整個重要性區塊於影像中保留原比例而非因擠壓使變形。
本論文提出一種基於影像切縫(Seam Carving)的調整方法，結合影像梯度(Gradient)測量、HSV色彩空間、皮膚偵測方法計算顯著性影像(Saliency Map)，並作為影像尺寸調整的依據。於迭代(Iteration)過程中使用小波轉換(Wavelet Decompose)分析，小波係數變異的幅度等同於影像結構破壞的程度，以此關聯變換使用均勻性縮放方法，以門檻值限制資訊遺失的程度。
本論文的實驗設計對影像的調整極限比較及使用性評估來證明提出之方法，於特定影像種類更能簡單和有效地，基於影像內容調適影像於特定長寬比，也維持影像原有的品質。

關鍵字

基於影像內容 ； 影像尺寸調整技術 ； 重要性區塊 ； 影像切縫 ； 顯著性影像 ； 小波轉換

並列摘要

Content-Aware Image Retargeting, differ from the traditional resize method which recognition of the Region-of-Interest(ROI) based on Content-Aware. After change the aspect ratio while preserving visually prominent whole regions rather than distributes the distortion.
In this article, we will propose a resize method that is based on the Seam Carving. The method base on an Saliency Map that combine the result of color gradient and HSV color space and skin detection to determine visual importance of pixels. Using the wavelet coefficient range of variation with a threshold to avoid seriously artifacts, and switch the resizing method from our method to scaling for preventing the shape distortion.
In our experiment show the image retargeting result with limitation, and analysis the practical feasibility to prove our method preserves the important contents better than existing methods.

並列關鍵字

Content-Aware ； Image Retargeting ； Region-of-Interest ； Seam Carving ； Saliency Map ； Wavelet Transform

參考文獻

[2] C. K. Laurent Itti, and Ernst Niebur, "A Model of Saliency-Based Visual Attention for Rapid Scene Analsis," IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, vol. 20, pp. 1254-1259, November 1998.

[3] F. L. Yanwen Guo, Jian Shi, "Image Retargeting Using Mesh Parametrization," IEEE Transactions on Multimedia, 2009.

[6] H. Liu, et al., "Region-based visual attention analysis with its application in image browsing on small displays," presented at the Proceedings of the 15th international conference on Multimedia, Augsburg, Germany, 2007.

[10] K.-S. C. Jong-Woo Han, Tae-Shick Wang, Sung-Hyun Cheon, Sung-Jea Ko, "Improved Seam Carving using a modified energy function based on wavelet decomposition," presented at the Consumer Electronics, 2009. ISCE '09. IEEE 13th International Symposiumon, 2009.

[13] M. Rubinstein, et al., "Improved Seam Carving for video retargeting," presented at the ACM SIGGRAPH 2008 papers, Los Angeles, California, 2008.