Age Estimation using Fractal Dimension-based Wrinkle Texture

[[abstract]]在電腦圖學以及機器視覺部分，運用人臉圖像進行年齡估測試一個很流行的主題，是一項感興趣的研究。 在本篇論文中，我們使用碎形維度作為年齡估測中的依據，並且利用圖像的前處理，提升年齡估測的效果的。系統架構上可分前處理模組、碎形處理模組與判斷識別模組三個模組。將資料庫的影像輸入到本系統，透過前處理模組我們利用Haar-like人臉偵測技術框出人臉之後，再藉由人臉器官位置定位出皺紋區域並進行前處理。接著對皺紋進行碎形維度的計算，如DBC、Shifting-DBC、Scanning BC。最後根據碎形維度的變化利用轉換函數估測出人臉的年齡。由實驗結果顯示出，估測準確度到達72%。 其在實驗中運用了FG-NET資料庫作為實驗用資料庫，FG-NET資料庫記錄了與圖片相關的年齡與成長的資料，是年齡估測的好幫手。[[abstract]]Derived from rapid advances in computer graphics and machine vision, computer-based age estimation via facial wrinkle characteristics has become a prevalent topic as well as an interesting study. In this paper, we use the fractal dimension as the basis for estimating the age, and pre-processing of the image to improve the age estimation. This system contains one pre-processing module, one fractal process module and one determination module. The images are input to this system. The pre-processing module detects the face area via Haar-like method. The wrinkle area is positioned by locating organs on the the face and pre-treated. Then, fractal dimension calculation on the wrinkle area is proceeded by means of DBC, Shifting-DBC, and Scanning BC. Finally, based on the changes in the of fractal dimension, conversion functions are used for estimating the age of the human face. The experimental results demonstrate that the estimation accuracy reaches 72%. In this thesis, all experimental results are tested on FG-NET database. Each image in this database is recorded with ages and growth background of a person. The database does help to test our age estimation system.[[note]]碩