Research On Low-resolution Face Recognition Based On Representation Learning

Face recognition technology has attracted a lot of attention from the academic community due to its wide application prospects,In the real world,Most face images are of poor quality,Different resolutions,It contains less information,and it is sometimes difficult to complete recognition under the influence of uncertain factors such as lighting and posture.This article will proceed from the problem of local image blocks and study low-resolution face recognition.Therefore,the main contents and innovations of this article are as follows:(1)This paper proposes a low-resolution face recognition method based on feature representation set,Obtain image blocks at each pixel position of the test and training samples,Use the Schatten-p norm instead of the kernel norm in previous studies,Good experimental results are obtained under the occluded AR face database and the Extended Yale B face database.(2)This paper proposes a cross-modal low-resolution face recognition method based on feature learning,and by projecting a face image into a common feature space,And used a discriminatory representation learning method,Thus,cross-resolution face image recognition is realized.Experimental results on the CUHK sketch database and AR database show that the method proposed in this paper is effective and superior for some advanced face recognition methods.(3)Because low-resolution face images have less information available,In order to improve the calculation efficiency,In this paper,we use face feature point detection technology to extract key pixels of face,and propose a low-resolution face recognition method with convolutional neural network features.For each image patch,Get the feature vector through the learning of the neural network,The weighted sparse coding regular regression method is used to linearly represent the feature vector of the test image block and the feature vector of the training image block,Finally complete the classification of each test image block.The experimental results obtained in AR face database and Extended Yale B face database both verify the effectiveness of the algorithm.