Pengecaman wajah menggunakan kaedah graf terhubung
Human face recognition is very important to the national’s security. In Computer Vision, human face recognition is one of the important areas in the digital image processing. Face identifications systems are categorized into two general categories which are front-view or full face and side-view or...
محفوظ في:
المؤلف الرئيسي: | |
---|---|
التنسيق: | أطروحة |
اللغة: | English |
منشور في: |
2006
|
الموضوعات: | |
الوصول للمادة أونلاين: | http://eprints.utm.my/id/eprint/1287/1/IsmailMatAminPFS2006.pdf |
الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
الملخص: | Human face recognition is very important to the national’s security. In Computer Vision, human face recognition is one of the important areas in the digital image processing. Face identifications systems are categorized into two general categories which are front-view or full face and side-view or profile. Front-view of human faces will be focused in this thesis. The problems that aroused are such as a large size storage and long time processing. The features in face images are extracted in the proposed method using connected graph, which uses pixel distance and proportion. ABC filter is proposed in the pre-processing stage to ensure the edges of eye, nose and mouth is clear. This technique is comparable to the invariant moment concerning the size and orientation. In the recognition process, the features are mapped on the space of feature reference where the Euclidean distance and statistical distance between face image reference and face image test are used as to determine a successful recognition. The result of recognition rate on Feret database with 10 reference image and 70 testing image collect randomly. The rate of recognition result for connected graph is 84.3 percent and invariant moment is 81.4 percent. Therefore, the result of this research proves that the algorithm proposed is equivalent to the invariant moment technique which has invariant towards the size and orientation but perform better in recognition rate |
---|