基於影像美感計算之智慧型圖文合成系統

An Intelligent Composition System for Text-overlaid Images based on Computational Aesthetics

賴建穎(Chien-Yin Lai)

指導教授 ： 洪政欣

國立暨南國際大學/科技學院/資訊工程學系/博士(2009年)

摘要

許多的事證已顯示美學計算在各個領域中的重要性；對於視覺介面上美感的研究，以往都是藉由定義許多的定性原則來進行分析，而沒有一套最佳的定量方法。在這個研究中所考量的視覺介面被稱為「圖文影像」，它通常是由少數的文字放置在一張大尺寸的影像上所構成。在很多現實生活中的應用，都需要將影像與文字做即時的合併，例如：線上電子賀卡(明信片)服務。在這樣的情況下，需要有一個工具來評量這些圖文影像的美感，也就是一個即時的圖文合成系統能夠被實現。

這篇論文首先針對圖文影像的「平衡」以及「對稱」兩種特性，提出基於「視覺重量原則」以及「影像分割技術」的量化模型，並藉著兩個實驗來研究平衡與對稱在圖文影像上美感的效用。實驗結果發現圖文影像的美感與平衡間有一個比例關係的存在。基於上述結果，我們實作出一個智慧型系統能夠計算出最佳的文字位置並且自動地構成圖文影像。一個基於「粒子群最佳化演算法」來計算最佳文字位置的最佳化引擎也已經被發展出。在論文的最後一個部份，提出一個試驗性研究，以「影像特徵圖」來計算具有複雜場景圖文影像的平衡特性。在影像中每個物件的視覺重量都是藉由影像特徵圖來計算評估。藉由一個實驗來研究已計算出的平衡是否與圖文影像的美感存在著關係。實驗結果呈現平衡與整體美感有一個正向關係，而這個結果剛好與使用影像分割技術方法所得到的結果一致。

關鍵字

計算美學 ； 介面設計 ； 組成特性 ； 影像分割 ； 粒子群最佳化 ； 影像特徵圖

並列摘要

Evidence in support of the importance of aesthetics in various aspects of computing has emerged recently. Research on visual interfaces with aesthetic considerations has traditionally defined visual aesthetics using various qualitative design principles but without knowing quantitatively where an optimal design space exists. The type of visual interface addressed in this study is the “text-overlaid image”, which usually consists of a large-size “background image” with a small number of texts overlaid on it. In many real-life applications such as an online greeting card (or postcard) service, there is a need to automatically overlay the texts on the pictorial image because the image and texts are randomly selected and composed together on-the-fly. In such contexts, it is crucial to have a tool to measure the aesthetic appeal of the resultant composed images.

This thesis first describes computational models based on principles of visual weights and image segmentation techniques to compute the balance and symmetry of text-overlaid images. Two experiments were conducted to investigate the effects of balance and symmetry on the aesthetic appeal of text-overlaid images. The experiments established a relationship between a higher averaged visual balance and the aesthetic appeal of text-overlaid images. Based on the above results, we have implemented an intelligent system that compute the optimal text position for automatically overlaying a paragraph of texts on a given background image. A computationally efficient algorithm based on Particle Swarm Optimization for calculating the optimal text position is developed. In the last part of the thesis, a pilot study for computing visual balance of text-overlaid images with complex scenes using image saliency map techniques is proposed. The visual weight of each object in a given image is estimated based on the image saliency map. An experiment was conducted to investigate the relationships between the computed values of balance and the aesthetic appeal of text-overlaid images. The experimental results show a strong positive relationship between averaged balance and overall aesthetic appeal. The finding well confirms to the results investigated by using the segmentation-based approach.

並列關鍵字

Computational aesthetics ； Interface design ； Compositional attributes ； Image segmentation ； Particle swarm optimization ； Image saliency map

參考文獻

Arnheim, R., 1974. Art and Visual Perception. University of California Press, Berkeley.

Arnheim, R., 1988. The Power of the Center. University of California Press, Berkeley.

Bauerly, M., Liu, Y., 2006. Computational modeling and experimental investigation of effects of compositional elements on interface and design aesthetics. International Journal of Human-Computer Studies 64(8), 670–682.

Bauerly, M., Liu, Y., 2008a. Effects of symmetry and grouping on interface and design aesthetics. International Journal of Human-Computer Interaction 24(3), 275–287.

Bauerly, M., Liu, Y., 2008b. Evaluation and improvement of interface aesthetics with an interactive genetic algorithm (IGA). International Journal of Human-Computer Interaction 25(2), 155–166.