Hyperspectral Face Recognition by Texture Feature Extraction Using Hybrid Wavelets Type I & Ii and Kekre Wavelet Transform

Abstract

Distinct from other normal images, hyperspectral images consist of huge data, but its analysis requires deep understanding of exactly of what properties are to be measured and their relation to the actual measurements made by Hyperspectral sensor. With the enhancement and obtainability of such Hyperspectral face data has increased the possibility of building such systems. The emphasis of this research is to recognize hyperspectral face images and compare the results obtained from the implementation of Multimodal system. The implementation of this System is based on three main texture feature extraction techniques. Hybrid Wavelet Type I (HWI), Hybrid Wavelet Type II (HWII) and Kekre Wavelet (KW) are used to generate these Multi Resolution Analysis based Feature Vectors. These Feature vector databases are then put through Intra Class testing And Inter Class Testing and their performance is evaluated and compared with existing systems. Proposed system has given up to 77% of EER for TAR-TRR (PI) for the multimodal implementation.

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

“…The results suggest that the removed noisy bands contain usefulinformation and 3D‐DWT has an advantage of extracting them. Table 3 shows significant improvement of recognitionover [7], after the proposed 3D‐GE is appliedto third scenario for frontal, right and left 94.16% ± 1.23, 78.94% ± 2.19 and83.2% ± 1.67 by using k‐NN classifier and 94.43% ± 2.1, 83.35% ± 2.3 and87.97% ± 1.6 by using CRC. The experimental results of fourth testing scenario areshown in Table 4.…”

“…Furthermore, distinct personal patterns originating from tissues, blood and organ structure can be captured using hyperspectral imaging. By employing the application of hyperspectral imaging, difficulties encountered in visible light‐based face recognition systems, such as the variance in orientation, illumination or expressions can be minimised [7]. Fig.…”

“…Shen and Zheng [1] employed 3D Gabor wavelet with changing central frequencies, in order to extract features in spatio‐spectral domain. Vartak and Bharadi [7] have introduced Kekre wavelet (KW), hybrid wavelets type I and type II (HWI and HWII) to generate multi‐resolution feature vectors. They have employed intra‐ and inter‐class test scenarios for the performance analysis of feature vector datasets with existing systems.…”

Robust and adaptive algorithm for hyperspectral palmprint region of interest extraction

“…Currently, there has been a trend to use hyperspectral imaging techniques for biometrics such as face [10,11] and palmprint [12][13][14][15]. In hyperspectral imaging, the recorded spectrum has a fine wavelength resolution and covers a wide range of wavelengths [16].…”