Image Composition Optimization based on Aesthetics Rules

• Main Authors: Cai-Chen Lin, 林采辰
• Other Authors: 李明穗
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/89565910241807281252

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 100 === This work presents a system to improve an image’s aesthetics by shifting the main object''s position to change image composition. The system is user-friendly. Without knowledge of any aesthetic rules, users obtain a more aesthetically pleasing result needing only to loosely draw a rectangle and make some brushes to provide the information of main object''s location. The system wants to provide a convenient editing tool to the rapidly growing community of amateur photographers in this modern society. A cost function is designed to find the main object''s final position by considering the rule of thirds, visual balance, and avoiding dissection lines in the background passing through facial regions. It also rejects the target which does not obey semantic constraints. The system starts by extracting the main object. Then the center of mass is located by producing the image''s saliency map. After applying the cost function, system arrives at the best target position of the main object. Furthermore, when the main object is too small resulting in an ambiguous image, or background ratio is improper, the system will fix it automatically and get the resultant image with more aesthetics. Not everyone in the rapidly growing populous has the expertise of a professional photographer’s with respect to scene selection and photographing, nor are they familiar with photograph-editing software as art professionals. For that very reason, a simple path towards post-editing and enhancing aesthetics of a photograph, and contributing to increasing its overall potential, is also one of the purposes of this research.