以特徵為主之物件修補技術

A feature-based object inpainting technique

蕭俊男(Chun-Nan Hsiao)

指導教授 ： 黃惠俞

國立虎尾科技大學/電機資訊學院/資訊工程研究所/碩士(2010年)

摘要

網路資訊的發達，利用網路傳播資訊已是相當普及且是獲得資訊的一項不可或缺之媒介，在網路上互相分享著各種資訊如旅遊的經驗、照片等。透過一些影像處理技術可將接收到之照片，將有興趣的物件進行移動、刪除之處理，創造出一張新的影像，達到想要的效果是一項相當有趣的研究議題。而本論文將利用樣本(exemplar)的影像修補技術及結合相關的影像處理，設計一處理程序對影像進行物件的移除及修補。透過物件擷取及影像修補，可以任意的移動或是移除影像中的物件，並可將移除物件後的影像進行修復進而展示無物件置於前的影像。
本論文研究是以特徵為主之物件修補技術，主要將前景與背景分離過後，針對擷取出的前景物件所遺留於背景上之空白區域進行修復，將此區域與背景相符合的進行填充處理。最後將物件移動及區域填充完後，則可獲得一張新的影像。本論文主要強調影像修補的部份，經由修補目標區域上各點之信賴度及亮度變異度的計算推演出各點優先權值，藉由排序修補優先權，進而找出最優先修補區塊，並利用影像中除目標區域外的可用資訊找尋出與該修補區塊最相似的資訊進行填補，而實驗證明經由以上程序處理後，本論文所提出之方法可獲得不錯的修補結果，並且在執行時間上有很大的改善。

關鍵字

樣本 ； 物件擷取 ； 影像修補

並列摘要

The development of internet advances more and more fast; the information given becomes very easy and fast by mean of internet. There are many shared media or advantage information, such as pictures or photos or messages, can download to further use by people. Owing to the pictures are easy to obtain by internet, human beings can use the image processing technology to modify these pictures that can further create a new and interesting picture. This is, human beings can re-make who desired the picture into a new picture or a virtual scene. For re-make picture scheme, it is very interesting topic and its applications can widely use in many fields, such as virtual reality, object deletion, etc. Based on this, we will develop an inpainting method based on exemplar viewpoint to achieve the picture remake. The aim of this thesis is to move the objects in the region of interest and to repair the moved region.
In this thesis, a feature-based object inpainting technique has been proposed. The main phases consist of the foreground and background segmentation, and region inpainting. When the foreground object extracted, the rest of region can be repaired and constructed a new image. In the inpainting process, there are two major factors that are confidence and variation to determine the priority of the filling order of each point in target region. Based on these two factors, we design a series of procedures to perform the inpainting. The experimental results demonstrate that the proposed method can obtain good repaired result and visual quality, and the computational time can be improved.

並列關鍵字

Exemplar ； object extraction ； image inpainting

參考文獻

[1]J. Sun, L. Yuan, J. Jiam, and H. Y. Shum, “Image completion with structure propagation,” in Proc. of ACM SIGGRAPH, 2005, pp. 861–868.

[2]N. Komodakis and G. Tziritas, “Image completion using global optimization,” in Proc. of IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, 2006, pp. 442–452.

[3]M. Bertalmio, G. Sapiro, V. Caselles, and C. Ballestrt, “Image inpainting,” in Proc. ACM Conf. on Comp. Graphics.(SIGGRA-PH), 2000, pp. 417–424.

[4]M. Bertalmio, A. Bertozzi, and G. Sapiro, “Navier stokes, fluiddynamics and image and video inpainting,” in Proc. of IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, 2001, pp. 1335–1362.

[5]T. K. Shin and R. C. Chang, “Digital inpainting-survey and mulitiplayer image inpainting algorithms,” in Proc. of the 3th Int. Conf. on Information Technology and Applications, 2005, pp. 15–24.