Detecting Smile and Forgery in Digital Color Images

此份論文中提出了三項應用於數位影像處理的技巧, 其中我們設計了演算法及改良技巧, 並且得到了不錯的結果. 有些公開發表的文獻中有提到處理人臉表情的偵測, 他們大部分都使用了類神經網路技術, 而這也表示了相當的花時間. 論文第一章提出了對於偵測正面笑臉的一個新穎、簡單並且有效的演算法, 而此方法可運用成為數位相機中的一項獨特功能. 我們使用的方法中, 笑臉偵測技巧主要包含了三個用於分割臉部的步驟: 首先我們加強了皮膚影像中陰影區域的紋理, 之後追蹤面貌的局部化技巧被應用來找尋嘴巴的部分, 最後笑臉就可以被我們所提出的技巧所偵測到.文第二章提出了一項新方法, 可用於偵測數位彩色影像的偽造. 因為有很多有效的方式可以製造這些影像, 使得也有很多各式各樣的軟體工具可以用來偵測它們. 對每一個影像來說都是獨一無二的挑戰, 所以需要不同的工具來偵測. 就大部分的人造影像而言, 影像中的所有物體通常是很難保證能夠均勻的受光, 因此不太能夠有著均勻分布的陰影. 從這點切入, 我們發展了能夠發現位於偽造影像物體間陰影不均勻的方法. 我們的方法採用的是, 偵測由使用者切割的皮膚區域, 找出陰影和明亮的皮膚區域, 接著描繪出從皮膚明亮區域重心到皮膚陰影區域重心的向量.部偵測被認為是物件種類偵測的一種, 也就是說主要工作是找到影像中屬於某些特定種類的所有物件的位置和大小, 比如說人類, 建築物, 或是車輛. 臉型偵測不僅僅用於臉型辨識的一部分, 更可應用於影像監視系統, 人機互動界面, 影像資料庫管理, 甚至是最近流行的具備可自動對焦功能的數位相機. 論文第三章部分, 我們介紹了一些簡單的臉部偵測技巧, 一種可用來決定在數位影像中人臉的位置而忽略其他區域的演算法. 演算法主要是基於切割影像為皮膚和非皮膚的區域, 然後利用臉貌特徵, 我們就可以決定該區域是否為臉的部分.This thesis proposes three techniques for digital image processing, in which we designed algorithms, enhanced techniques and came up with good results. There are open literature dealt with detecting human face expressions and most of them are done using neural network technique that means very time consuming. First chapter proposes a new simple and an efficient algorithm for smile detection in frontal face; in which it could be exploited as one unique function of camera. In our approach, the smile detection technique consists mainly of three steps used on segmented faces: the first one is by enhancing the lamination in shadow areas of the skin image, afterwards, localizing technique for tracing facial features is done to find the mouth area, and finally the smile is detected using our proposed technique.he second chapter proposes a new method for detecting forgery in digital color images, since there are many efficient ways to create these images; there are also variety tools to detect them where each image is a unique challenge and need different tools to detect them. In most of the artificial images, it is very hard to ensure that all objects are consistently illuminated and therefore the shadow is consistently distributed; from this point, we developed a method which discovers the shadow inconsistency among objects in digital forgery images, our algorithm is done by detecting skin regions from user segmented areas, then shadow and highlight skin regions are found, therefore, vectors are drawn from the center of gravity of the skin highlight region to the center of gravity of the skin shadow region in each skin segmented areas.ace detection is considered as a case from object-class detection, in which the task is to find the locations and sizes of all objects in an image that belong to a given class, such as peoples, building, and cars. Face detection is not only used as a part of a facial recognition system, but also used in video surveillance, human computer interface and image database management, furthermore, some recent digital cameras use face detection for auto-focusing task. At the third chapter, we introduced simple face detection technique, which is an algorithm to determine the locations of human faces in digital images and ignoring anything else, it is based on segmenting the input image into skin and non-skin regions, then by using facial feature, we decide whether this skin region would be a part of the face or not.Abstract icknowledgements vihapter one 1fficient Smile Detection Using Second Order Normalized Spatial Moment 1.1) Introduction: 1.2) Color spaces 2.2.1) Gray Space 3.2.2) RGB Color Space 3.2.3) YCbCr Color Space 5.3) Simple techniques to detect shadow in real images 7.3.1) Shadow removal from real images based on the color cluster in RGB color space 7.3.2) Shadow removal from a real image based on shadow density 9.4) Lumination enhancement in facial shadow areas 11.4.1) Skin area detection 12.4.2) Lumination Enhancement in Shadow area: 15.5) Mouth localization using geometric generic face model: 17.6) Smile detection using second order normalized spatial moment 20.7) Experimental results 26hapter Two 30etecting Digital Forgery Based on Shadow Inconsistency in Lighting 30.1) Introduction: 30.2) Shadow 31.2.1) Variation with time 32.2.2) Non-point light source 32.3) Some techniques for detecting forgery image 33.3.1) Duplication 34.3.2) Camera Finger prints 35.3.3) Specular highlight 37.3.4) In a different light 39.4) Our algorithm for detecting digital forgery based on shadow inconsistency in lighting 41.5) Experimental results 45hapter three 49 simple technique for Face Detection 49.1) Introduction 49.2) Face detection algorithm 49.3) Experimental result 51hapter Four 53onclusion and Future Work 53eferences 5