材質分割與分類於SVG漫畫壓縮之應用

Texture segmentation and classification for SVG Comic Compression

林琮翰(Tsung-Han Lin)

指導教授 ： 張瑞益

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2014年)

https://doi.org/10.6342/NTU.2014.10641

摘要

在攜帶式裝置上，點陣格式的漫畫在縮放時會導致漫畫的品質降低。雖然將漫畫以轉換為向量形式可以避免此問題，向量漫畫有較大的檔案大小及較慢的顯像速度。我們提出一個以SVG格式為基礎的壓縮方法，能在將點陣漫畫轉換為SVG時，降低轉換後的檔案大小及顯像時間。我們先使用材質分割技術將漫畫分為材質與非材質區域，接著在將圖像轉換為SVG時將材質區域以SVG中的元素儲存來達到效果。在材質分割時我們使用CSGV(Composite sub-band Gradient Vector)作為特徵值，以SVM(Support Vector Machine)分類漫畫中的每個區域。再使用基於KL (Kullback-Leibler)距離及Split-Bregman方法進行演算的動態輪廓模組來增加分割準確率。我們對此方法以若干合成的漫畫進行實驗。實驗結果顯示此方法能讓向量漫畫在攜帶型裝置上達到更高的品質與效能。處理過的SVG圖檔，平均能減少55.3%的檔案大小及61.37%的顯示時間。此外，這方法也同時能使用在內含複數材質的漫畫上。

關鍵字

圖像分割 ； 可變向量圖形 ； 支持向量機器 ； 機器學習 ； 動態輪廓模型 ； 影像壓縮

並列摘要

In portable device, scaling raster manga would result in reduced manga quality. Although converting manga into vector format could avoid this problem, vector manga has larger file size and slower rendering speed. We present a compression method based on SVG format, which can reduce file size and rendering time when converting raster manga into SVG format. We first use texture segmentation techniques to partition manga into texture segments and non-texture segment, then we use element to store texture segments when converting manga. In image segmentation, we use Composite Sub-band Gradient Vector as texture descriptor and use Support Vector Machine to classify every area in manga. Then we use Active Contour Model, which based on KL (Kullback-Leibler) distance and Split-Bregman method, to enhance accuracy of segmentation. We conduct some experiments using several manga to test this method. Result shows this method can let vectorized manga have higher performance on portable device. In average, Segmentation accuracy is 93.3%, reduced file size is 55.3% and reduced rendering time is 61.37%. In addition, this method can also be applied on manga with multiple textures.

並列關鍵字

Image segmentation ； Scalable Vector Graphic (SVG) ； Support Vector Machine (SVM) ； Machine Learning ； Active Contour Model ； Image Compression

參考文獻

[4] K. Kawamura, Y. Yamamoto, and H. Watanabe, "Gradation approximation for vector based compression of comic images," in Image Processing, 2005. ICIP 2005. IEEE International Conference on, 2005, pp. III-489-92.

[5] P. Huang, S. Dai, and P. Lin, "Texture image retrieval and image segmentation using composite sub-band gradient vectors," Journal of Visual Communication and Image Representation, vol. 17, pp. 947-957, 2006.

[6] V. Manian and R. Vasquez, "Scaled and rotated texture classification using a class of basis functions," Pattern Recognition, vol. 31, pp. 1937-1948, 1998.

[8] B. Scholkopf, J. C. Platt, J. Shawe-Taylor, A. J. Smola, and R. C. Williamson, "Estimating the support of a high-dimensional distribution," Neural computation, vol. 13, pp. 1443-1471, 2001.

[9] N. Houhou, J.-P. Thiran, and X. Bresson, "Fast Texture Segmentation Based on Semi-Local Region Descriptor and Active Contour," Numerical Mathematics: Theory, Methods & Applications, vol. 2, 2009.