Extraction of Content -Free Basal Image and Its Application in Face Recognition

Extraction of Content -Free Basal Image and Its Application in Face Recognition

연구 분야: Artificial Intelligence

학회: 2020 Chinese Control And Decision Conference (CCDC)

As a typical signal, an image can be composed of a series of basic signals, which are called basal images. In order to find the basic signals of a group of reconfigurable arbitrary images and optimize the face recognition algorithm, a feature-based basal image extraction method is proposed and applied to face recognition, so that the basal image can be extracted from any set of images and the performance of face recognition algorithm to local changes such as illumination and expression can be improved. Feature extraction algorithm is used to decompose a series of basal images from the training set image, the algorithm flow of basal image decomposition and extraction is expounded, the projection coefficient is obtained by projecting the face image into the space constituted by the basal image, the distance between the test sample and the training sample coefficient matrix is calculated, and the minimum distance classifier is used for classification. The experimental results show that the basal image obtained from any image set can be used for face recognition, and the basal image have content independence. At the same time, this method is better than the traditional feature face recognition method in light, expression, wearing accessories and other local changes.