在這篇論文中，我們提出一個建築物影像的污漬清除系統，其中污漬包含影像牆壁上的水漬、裂縫…等，為了達到污漬清除，我們結合了材質合成、Poisson Image Editing等技術，使得修復結果貌似真實，我們也利用區域亮度的統計資訊配合影像金字塔來達到污漬偵測的功能，結合污漬偵測和污漬清除這兩大功能，我們實作出一個可自動化的建築物影像污漬清除系統，以下簡單描述一下我們系統的步驟。首先，使用者載入一張含有污漬的建築物影像，系統利用影像的亮度統計資訊藉由迭代的方式反覆不斷地切割影像來進行區域亮度門檻篩選，配合影像金字塔的應用在多重解析度下自動偵測出可能是污漬的區域；接著，藉由系統在尋找污漬區域時所產生的DirtyDegreeMap來搜尋污漬周圍相對乾淨的材質區域，進行材質合成取代污漬區域；但經過材質合成後取代污漬的部分可能會因為亮度、顏色的差異和原來污漬周圍的區域產生細縫，所以我們可以選擇性地使用Perez提出Poisson Image Editing的方式進行無縫貼圖，使材質合成後的乾淨材質可以無縫的跟週遭區域結合；最後，影像中建築物的外觀變得沒有污漬，使得建築物變得更新且完整。

In this thesis, a novel image restoration technique is proposed for repairing building facades with smirches in a 2D image. In order to clean the smirches, we combine texture synthesis and Poisson image editing techniques for the repairing process. For smirch detection, we make use of the local brightness information and apply an image pyramid approach. Through an integration of the aforementioned methods, our semi-automatic smirch cleaning system proceeds as follows. First, given a 2D building image with smirches, our system iteratively splits the image into many overlapped sub-images in a multi-resolution manner, and performs local brightness thresholding processes to detect dirty regions respectively. Second, for each dirty region, we search for its surrounding cleaner region for the re-synthesis purpose. Finally, we utilize Poisson image editing to seamlessly replace a dirty region by its cleaner counterpart. As a result, our system can clean the smirches on the building facades in a 2D image.

[1] EFROS A. A., LEUNG T. K.: Texture Synthesis by Nonparametric Sampling. In International Conference on Computer Vision (1999), pp. 1033-1038.

[2] WEI L., LEVOY M.: Fast Texture Synthesis Using Tree-Structured Vector Quantization. In SIGGRAPH '2000 (2000), pp. 479-488.

[3] ASHIKHMIN M.: Synthesizing natural textures. ACM Symposium on Interactive 3D Graphics, 2001, pp. 217-226.

[4] HERTZMANN A., JACOBS C., OLIVER N., CURLESS B., SALESIN D.: Image analogies. In SIGGRAPH '2001 (2001), pp. 327-340.

[5] HARRISON P.: A Non-hierarchical Procedure for Re-synthesis of Complex Textures. In Proceedings of Central Europe on Computer Graphics, Visualization and Computer Vision, pp. 190-197, 2001.

[6] ZELINKA S., GARLAND M.: Towards real-time texture synthesis with the jump map. In Proceedings of the Thirteenth Eurographics Workshop on Rendering Techniques, 2002, pp. 99-104.

[7] XU Y., GUO B., SHUM H.-Y.: Chaos mosaic: Fast and memory efficient texture synthesis. Microsoft Res. Tech. Rep. MSR-TR-2000-32, April 2000.

[8] EFROS A. A., FREEMAN W. T.: Image Quilting for Texture Synthesis and Transfer. In SIGGRAPH '2001 (2001), pp. 341-346.

[9] LIANG L., LIU C., XU Y., GUO B., SHUM H.-Y.: Real-time texture synthesis using patch-based sampling. ACM Trans. Graphics, Vol. 20, No. 3, 2001, pp. 127-150.

[10] NEALEN A., ALEXA M.: Hybrid Texture Synthesis. In Proceedings of Eurographics Symposium on Rendering, 2003, pp. 97-105.

[11] KWATRA V., SCHÖDL A., ESSA I., TURK G.,BOBICK A.: Graphcut textures: Image and video synthesis using graph cuts. In SIGGRAPH '2003 (2003), pp. 277–286.

[12] KWATRA, V., ESSA, I., BOBICK, A., AND KWATRA, N.: Texture optimization for example-based synthesis. In SIGGRAPH '2005 (2005), pp. 795-802.

[13] EISENACHER C., LEFEBVRE S., STAMMINGER M.: Texture synthesis from photographs. In Proceedings of the Eurographics conference (2008).

[14] LU J., DORSEY J., RUSHMEIER H.: Dominant Texture and Diffusion Distance Manifolds. In Proceedings of Eurographics 2009

[15] BERTALMIO M., SAPIRO G., CASELLES V., BALLESTER C.: Image Inpainting. In SIGGRAPH '2000 (2000).

[16] OLIVEIRA M. M., BOWEN B., MCKENNA R., CHANG Y.: Fast Digital Image Inpainting. In VIIP (International Conference on Visualization, Imaging and Image Processing) '2001 (2001).

[17] YAMAUCHI H., HABER J., SEIDEL H.P.: Image restoration using multiresolution texture synthesis and image inpainting. In Proceedings of Computer Graphics International (CGI), pages 120-125, 2003.

[18] BERTALMIO M., VESE M., SAPIRO G., OSHER S.: Simultaneous structure and texture image inpainting. IEEE Transactions on Image Processing(2003), vol. 12, pp. 882–889.

[19] DRORI I., COHEN-OR D., YESHURUN H.: Fragment-based Image Completion. In SIGGRAPH '2003 (2003), pp. 303-312.

[20] SUN J., YUAN L., JIA J., SHUM H.: Image Completion with Structural Propagation. In SIGGRAPH '2005 (2005), pp. 861-868.

[21] XUE S., WANG J., TONG X., DAI Q., GUO B.: Image-based material weathering. Computer Graphics Forum 27, 2 (Apr. 2008), 617–626.

[22] HE L., CHAO Y., SUZUKI K.: A Run-Based Two-Scan Labeling Algorithm. IEEE Transactions on Image Process 17, 5 (2008), pp. 749-756.

[23] PORIKLI F., KOCAK T.: Fast Distance Transform Computation Using Dual Scan Line Propagation. In SPIE Conference Real-Time Image Processing, Vol. 6496, February 2007

[24] PEREZ P., GANGNET M., BLAKE A.: Poisson Image Editing. In SIGGRAPH '2003 (2003), pp. 313-318.