運用標籤賽局於數位影像之主體重新對焦

Refocusing on the Object by a Labeling Game for Digital Images

劉心如(Hsin-Ju Liu)

指導教授 ： 張隆紋

國立清華大學/電機資訊學院/資訊系統與應用研究所/碩士(2012年)

摘要

在攝影取像時，場景中通常會存在多重目標物，對相機而言，這些目標物位於不同的深度上，攝影者選擇一個深度作為對焦的依據，用景深來呈現的想要的情境。而一張相片所呈現的視覺品質，受限於相機鏡頭的種類與拍攝者的攝影技術。若拍攝時對焦不準確，便無法呈現出拍攝者的意念及感覺，導致它成為一張失敗的影像。為了解決這個問題，本文提出一個概念，讓攝影者在拍攝時不用考慮對焦的問題，只需隨意拍攝一張全對焦影像，再透過影像處理事後對焦至想要的目標或區域即可。
利用有限的影像資訊，我們提出一個以物件為基礎事後對焦架構，透過一個標籤賽局，來辨認出想要重新對焦的物件，透過空間濾波器使之在圖像中更為突出，並使背景模糊化模擬失焦的現象，進而達到景深設計的效果，提升整體圖像的視覺傳達品質。

關鍵字

對焦 ； 影像分割 ； 標籤

並列摘要

Generally there are more than one object that locate at different depth in a scene of a photographic image. Photographers select a depth as the basis of focusing to present the desired scenarios. However, the visual quality of a photo is limited by the type of camera lens and the skills of the photographer. It cannot present the photographer’s feeling and lead to a failure if we cannot focus on the wanted object properly. To solve this problem, this thesis proposes a concept that photographers don’t need to consider the focusing problem when shooting but refocus on it after computer processing.
With limited image information, we propose an object-based refocusing framework to refocus on the wanted object. Through a image labeling game algorithm, we correctly identify the object which we want to refocus and make it more prominent in the image by using a spatial filter. Similarly, we simulate the out-of-focus effect by blurring the background. After these processes, we retrieve a correct focused image that can present the desired depth of field and improve the quality of visual communication.

並列關鍵字

focusing ； segmentation ； labeling

參考文獻

[6] Mortensen, E. N., and Barrett, W. A. 1998. Interactive segmentation with intelligent scissors. Graphical Models and Image Processing 60, 5, 349–384.

[7] Boykov, Y., and Jolly, M.-P. 2001. Interactive graph cuts for optimal boundary and region segmentation of objects in n-d images. In Proc. Of the International Conference on Computer Vision, vol. 1, 105–112.

[8] Rother, C., Kolmogorov, V., and Blake, A. 2004. Grabcut – interactive object extraction using iterated graph cuts. Proc. ACM Siggraph.

[11] S. Yu and M. Berthod, A game strategy approach for image labeling, Computer Vision and Image Understanding, vol. 61, No. 1, January, 1995, 32-37.

[12] Guo dong Guo, Shan Yu and Song de Ma, An Image Labeling Algorithm Based on Cooperative Game Theory, Signal Processing Proceedings, vol.2, 1998,978 – 981.