Boosting and its application in face recognition and image retrieval

Abstract

There are many studies on the application of boosting in image processing, such as face recognition, face detection, image retrieval, and so on. By using an appropriate classifier, the accuracy of classification and computation in time are improved by applying boosting algorithm. In this dissertation, I have investigated the AdaBoost algorithm and conducted the experiments for feature selection and classification. It has be demonstrated that during the feature selection process by using AdaBoost algorithm, the strong classifier can be formed in linear combination of weak classifiers. The experimental results show that Adaboost algorithm is effective in feature selection. It also shows that features selected with lowest error of misclassification will be different if we choose different positive samples and negative samples. The importance of each feature for the sample images will be different too if the selected samples are different. I have studied the face recognition and image retrieval. The application of Ad-aBoost algorithm on feature selection for face recognition and image retrieval have been discussed too.