Multimodal low resolution face recognition using SVD
N. R. Roshna, S. Naveen · 2017
Face recognition is effective when input low resolutions are of high resolution (HR). Face images captured by surveillance cameras are of low resolution (LR) uncontrolled pose and illumination etc. The main objective is to find maximum accuracy in low resolution face recognition, for more number of test images. Here, set a LR and HR face images as reference images of each subject. Then find the error between LR test image and reference images using Principle Component Analysis(PCA). To find the error between LR test image and HR reference images, first perform the Singular Value Decomposition(SVD)of the test image and then do the bicubic interpolation technique, to make the size of the LR test image as same as that of HR reference images. The entire algorithm will be done for both texture and depth images. Then merge the error between both texture and depth images. Then find the recognition accuracy.