周遵儒Chou, Tzren-Ru陳仕杰Chen, Shih-Chieh2025-12-092025-08-122025https://etds.lib.ntnu.edu.tw/thesis/detail/0c45d54dacb668cdf095d8d439a2a8f4/http://rportal.lib.ntnu.edu.tw/handle/20.500.12235/125365隨著人工智慧與影像識別技術的迅速發展，設計領域逐漸探索如何將 AI 應用於構圖分析與視覺學習中，以突破傳統以經驗與直覺為主的構圖教學侷限。本研究以人物雜誌封面為切入點，嘗試建立一種客觀、量化且具擴充性的構圖分析模式，透過 AI 技術輔助視覺風格辨識與文化偏好觀察。本研究運用 MediaPipe Solutions 擷取封面中人物臉部與身體關鍵點座標，建立對應三分法、黃金比例與畫面中線等構圖法則的數據化量測流程，並以《VOGUE》雜誌為研究樣本，橫跨亞洲、美洲與歐洲三大文化區域進行封面圖像蒐集與構圖類型分類。研究以視覺焦點落點、臉部面積比例與水平、垂直構圖對應位置為主要分析指標，並輔以統計視覺化圖表呈現各地區的構圖風格偏好與分布特徵。結果顯示：亞洲樣本偏好臉部比例較大且視覺焦點集中於畫面中央與左側三分點，強調凝視感與人物主體性；歐洲樣本則傾向採用臉部比例較小、偏右構圖與空間留白，呈現古典敘事與藝術氛圍；美洲則展現出構圖自由與多元性，兼具視覺中心與邊緣落點的分布，呼應品牌多樣性與文化包容。這些跨文化的構圖變異，亦對應了 Face-ism 理論所揭示的「臉部比例影響觀感」之觀點，顯示人物主體性與視覺權威的展現與文化風格密切相關。綜而言之，本研究不僅提出一套整合MediaPipe Solutions技術與構圖理論的分析框架，有效量化人物構圖中的視覺焦點與比例特徵，亦從細微的構圖變化中觀察出文化設計趨向與視覺語法的潛在規律。此成果為 AI 輔助設計學習與圖像生成提供實證依據，未來有望成為具文化敏感度的視覺設計與美學訓練工具之基礎。With the rapid advancement of artificial intelligence and image recognition technologies, the application of AI in design composition learning has become a topic worthy of deeper exploration. Traditional composition instruction often relies on the intuition and experience of artists; however, in the face of large-scale digital images and increasingly diverse aesthetic demands, such an approach reveals limitations in both efficiency and consistency. This study proposes a more objective, quantifiable, and scalable method of learning composition through AI-assisted analysis.The research aims to develop a portrait composition analysis method based on facial recognition and pose estimation, utilizing MediaPipe Solutions to extract facial and body keypoints. This allows for a data-driven framework that corresponds to visual composition principles such as the rule of thirds, the golden ratio, and central alignment. The study focuses on analyzing proportional structures and focal points in magazine cover portraits, particularly within the context of design aesthetics.Taking VOGUE magazine covers as the primary sample, images were collected and categorized from three major cultural regions—Asia, Europe, and the Americas. Through the use of MediaPipe Solutions, facial keypoints were extracted from each image to analyze composition types and focal point locations. Statistical visualization was employed to present regional preferences in both vertical and horizontal compositional ratios, and further explored in relation to Face-ism theory.Results reveal distinct regional differences in composition: Asian samples tend to emphasize larger facial ratios and centralized or left-third alignment, enhancing gaze intensity and subject prominence; European samples favor smaller facial proportions, right-shifted compositions, and spatial breathing, creating a classical and narrative visual tone; American samples display greater compositional diversity and flexibility, balancing focal variation with brand inclusivity. These variations not only reflect aesthetic tendencies rooted in cultural visual contexts but also align with the Face-ism hypothesis that links facial prominence with viewer perception.In conclusion, this study introduces an analytical framework that integrates MediaPipe technology with classical composition theory, effectively quantifying visual focal points and proportional features in portrait design. By identifying compositional differences through subtle visual variations, the study offers insights into cultural design trends and underlying visual syntax, laying a foundation for future AI-assisted design applications and aesthetic modeling across diverse visual domains.AI人臉辨識身體座標比例特徵《VOGUE》封面構圖方法AI Facial RecognitionMediaPipe SolutionsBody KeypointsComposition RatiosVOGUE CoversVisual FocusCross-Cultural Comparison學術論文