人臉識別因其在安全驗證系統、信用卡驗證、醫學、檔案管理、視頻會議、人機交互、系統公安(罪犯識別等)等方面的巨大應用前景而越來越成為當前模式識別和人工智能領域的一個研究熱點。
人臉識別的關鍵就是人臉圖像預處理,預處理效果的好壞直接關系著人臉識別的結果。本文研究了一個基于PCA人臉識別的人臉圖像預處理方法,采用PCA方法就是利用K-L變換和SVD得到正交基,然后根據主特征提取的原理,選取較大特征值對應的基向量。目前算法僅僅針對單人正面的圖像,有很大的局限性。
本文所采用的方法,除了對算法的優化外,更加注重圖像預處理的效果。
關鍵字:人臉識別;人臉預處理;光照補償;K-L變換;面部特征定位
Research and Implementation Of Image Pre-processing Algorithm Based on Face Recognition
Abstract
Face recognition is important in surveillance and security, telecommunications, digital libraries , video meeting, and human-computer intelligent interactions. It has been a research focus of pattern recognition and artificial intelligence.
The important of face recognition is face image preprocessing. The pretreatment directly effect the results of the face recognition. In this paper, we study and implement a face image preprocessing method based on PCA. PCA use K-L transform and SVD to get orthogonal basis, and then according to the principles of the main feature extraction to select the larger eigenvalues of the corresponding vector-based. The current algorithm is only for the single positive face image and it have limitation.
The method in this paper uses algorithm optimization and the effects of image preprocessing to achieve the performance of the algorithm.
Keyword:Face recognition;Face pretreatment;Light compensating;K-L transform;Characteristics of position