陳炳宇臺灣大學：資訊管理學研究所邱立榕Chiu, Li-JungLi-JungChiu2007-11-262018-06-292007-11-262018-06-292006http://ntur.lib.ntu.edu.tw//handle/246246/54168本研究的重點是呈現一套系統, 使用的演算法可以產生具有漫畫風格的照片排版 (comic-styled photo layout), 其中包括將將多張照片放在同一頁 (layouting) 、將照片依需要適時地裁切 (crop)、對話泡泡 (speech bubbles) 的放置。本系統所實作出的編輯層 (authoring layer) 可以讓使用者很簡便地自行輸入註解對話 (annotations), 排版層 (layout generation layer) 可以將照片適時擺放在不同大小的輸出版面。所提出的演算法使用了幾項技術, 包括人臉偵測 (face detection)、重點區域偵測 (region of interest detection)、對話泡泡放置區域之偵測 (speech bubble placement detection) 及退火演算法 (simulated annealing)。我們也定義了不同排版結果間的距離函式及用來做排版最佳化的目標函式。因此, 我們可以對使用者想要排版的照片們做最佳化, 包括整體性(integrity), 強調性(emphasis), 及一致性(unity)。 在我們的實驗中, 在兩種不同的紙張大小裡使用三組照片資料作實驗, 裡面包括中文及英文的註解對話。另一方面, 我們也拿現在市場上能使用的其他六種相簿解決方案來做比較, 評估的量尺有使用方便度 (ease of use)、照片故事清楚度 (clarity of the photo story) 及排版結果有趣度 (interesting-ness)。The research presents a system and an algorithm for producing comic-styled photo layout that includes layouting the photos on a page, cropping the photos and placing the speech bubbles onto the photo. The system is comprised of an authoring layer that allows the user to key-in the annotations easily and a layout generation layer that automatically lays outs the photos on papers of varying sizes. The algorithm employs different techniques that include: face detection, region of Interest (ROI) detection, speech bubble placement detection and Simulated Annealing. Distance between layouts is defined by the research as well as the objective function that optimizes the integrity, emphasis and unity of the set of photos provided by the user. Three sets of data with two languages on two different paper sizes were explored as the test bed of this research. Six other market-available solutions were benchmarked against the presented thesis on the area of ease of use, clarity of the photo story, and the interesting-ness of the layout (how interesting the layout is).1.0 Research Description 1 1.1 Overview of Current State of Technology 1 1.2 Research Objectives 2 1.2.1 General Objective 2 1.2.2 Specific Objectives 2 1.2 Scope and Limitations of the Research 3 1.3 Significance of the Research 3 2.0 Review of Related Literature 4 2.1 Photo Information Acquisition 4 2.1.1 Embedded Information Extraction 4 2.1.2 Assistive Tool for Photo Story Sharing 6 2.1.2.1 Face-to-face 6 2.1.2.2 Human-computer-human 11 2.1.2.2.1 Sharing of Photo Story 11 2.1.2.2.2 Photo Story Authoring 12 2.2 Photo Layout Optimization 19 2.2.1 Genetic Algorithm 19 2.2.2 Simulated Annealing 22 3.0 System Overview 24 3.1 Authoring Layer 24 3.2 Presentation Layer 25 3.2.1 Bubble placement 25 3.2.2 Automatic Cropping 25 3.2.3 Page Layout 25 4.0 Framework and Algorithm 27 4.1 Authoring 27 4.2 Cropping Area Detection 28 4.2.1 ROI Detection 28 4.2.1.1 Polar Transformation of Features 28 4.2.1.2 Obtaining Subspaces 28 4.2.1.3 Obtaining the Attention Score of the Subspaces 29 4.2.2 Bubble Placement 29 4.3 Simulate Annealing Layout Generator 31 4.3.1 Solution representation 31 4.3.2 Input 32 4.3.3 Simulated Annealing Setup 33 4.3.3.1 Neighborhood 33 4.3.3.2 Objective Function 35 4.3.3.2.1 Integrity 35 4.3.3.2.2 Emphasis 36 4.3.3.2.3 Unity 37 4.3.3.2 Annealing Schedule 37 4.3.3.2.1 Initial Temperature 37 4.3.3.2.2 Final Temperature 38 4.3.3.2.3 Freezing Function 38 4.3.3.2.4 Length of Markov Chains 38 5.0 Results 39 6.0 Evaluation 44 6.1 Method 44 6.2 Results 45 6.3 Analysis 46 7.0 Conclusion 48 Appendix A: Bibliography 49 Appendix B: Resource Persons 52 Appendix C: Curriculum Vitae 53 Appendix D: Evaluation Form 55 Appendix E: Participant’s Demographics 56 Appendix F: Result Set Two (640 x 960) 58 Appendix G: Result Set Three (640 x 960) 62 Appendix H: Result Set One in Smaller Page Size (320 x 480) 71 Appendix I: Result Set Two in Smaller Page Size (320 x 480) 73 Appendix J: Result Set Three in Smaller Page Size (320 x 480) 7567483060 bytesapplication/pdfen-US具有漫畫風格的照片排版看圖說故事退火演算法重點區域偵測裁切照片對話泡泡放置comic-styled photo layoutphoto storySimulated Annealingregion of interest (ROI) detectionimage crop area detectionspeech bubble placementotherhttp://ntur.lib.ntu.edu.tw/bitstream/246246/54168/1/ntu-95-R93725050-1.pdf