Hyperspectral face recognition with log‐polar Fourier features and collaborative representation based voting classifiers

Abstract

Hyperspectral imagery analysis has become a popular topic for improving face recognition accuracy. Nevertheless, it encounters difficulty in data acquisition, low signal‐to‐noise ratio, and high dimensionality. As a result, there exists a need to develop better algorithms in order to achieve higher classification rates. In this study, the authors propose a new method for hyperspectral face recognition with very competitive experimental results. Since there is a significant amount of noise in every spectral band, they reduce noise adaptively from each spectral band by using any image denoising method, e.g. block matching and 3D filtering. They then crop each face according to its eye coordinates so that translation invariance can be achieved. They conduct log‐polar transform to each cropped face image and extract 2D Fourier spectrum from them. In this way, the extracted features are approximately invariant to translation, rotation, and scaling. They use the collaborative representation‐based classifier with voting for hyperspectral face recognition. They perform some experiments to test the authors’ new method for hyperspectral face recognition with very promising results.

3D discrete wavelet transform‐based feature extraction for hyperspectral face recognition

“…First expected and observed results and then experiment's degrees of freedom (DF) are determined. The chisquare is calculated by (12), where 'o' is the observed value and 'e' is the expected value. The results for PolyU and CMU are 38.67 and 69.35, respectively…”

“…They generated spatial-spectral feature descriptor by applying a 3D histogram on the derivative pattern, which can be used to convert hyperspectral face images into vectorised representations. Chen et al [12] conducted log-polar transform to each hyperspectral face image and extracted 2D Fourier spectrum from them. In [13], Cho et al performed an automatic selection framework for the optimal alignment method to address inter-band misalignments by developing two different metrics which evaluate the alignment quality of the hyperspectral image cubes.…”

On the use of ear and profile faces for distinguishing identical twins and nontwins

“…In Mahmood, Ali, and Khan (2016), the adaptive boosting with linear discriminant analysis as weak learner, the PCA-based approach, and LBP-based approach were used to analyse the effect of poses and image resolutions on face recognition. Chen, Sun, and Xie (2016) proposed to extract features that are robust to translation, rotation, and scaling using log-polar Fourier for hyperspectral face recognition. Huang, Shan, Ardabilian, Wang, and Chen (2011) summarize local structures of images efficiently by comparing each pixel with its neighbouring pixels.…”

Hyperspectral face recognition via feature extraction and CRC‐based classifier

“…Zhang et al [16] found out that it was the collaborative representation but not the sparse representation that improves face recognition results. In addition, Chen et al [4] introduce both partial voting (P:) and full voting (F:) based on the CRC classifier. In this work, we use the same voting techniques as [4] for hyperspectral face recognition.…”

“…They achieved state-of-the-art classification rates for three hyperspectral face databases (HSFD). Chen et al [4,5] worked on hyperspectral face recognition using log-polar Fourier features and collaborative representation-based voting classifiers (CRC). They found that CRC-based voting techniques are better than CRC alone for hyperspectral face recognition.…”