Search

A frequency‐based image deblurring method is proposed. Unlike the traditional ways of replicating target images, a periodic point spread function (PSF) is constructed. By investigating the periodic PSF under the Fourier domain, it has been found that noises at different frequencies produce significantly varying impacts on the reconstruction quality. Therefore, the Fourier slice theorem is utilised to select partial Fourier samples and to avoid the noises that can severely damage performance. A sparse‐inducing method is then developed to recover a clear image from the selected Fourier samples. [ABSTRACT FROM AUTHOR]

The sparse feature sphere centroid (SFSC) classifier for face recognition is proposed. SFSC is based on nearest feature plane (NFP), sparse representation classification (SRC) and nearest feature centres (NFC), and it contains two stages. In the first stage, the SFSC classifier computes the feature sphere centroid metric. Then, SFSC obtains the sparse coefficients by solving an L1‐norm minimisation problem and uses the sparse coefficients to calculate the weighted feature sphere centroid distance, which will be utilised for classification. Experiments on the Georgia Tech (GT) face database and AR face database were conducted to evaluate the proposed classifier. The experimental results show that the proposed classifier yields better recognition rate over competing classifiers such as NFC, NFP and SRC. [ABSTRACT FROM AUTHOR]