000			04388ntm a2200373 i 4500
999			_c91072 _d91078
003			MY-KuUP
005			20251125105419.0
006			t||||fr|||| 000 0
007			ta
008			191105t20182018my a|||fram|| 000 0 eng d
020			_aTHE0008192(Local) _qhardback
040			_aUMP _beng _cUMP _erda
090			_aFKEE .S64 2018 r Thesis
100	1		_aSoh, Shi Chuan, _eauthor.
245	1	0	_aMultispectral palm vein image fusion for contactless palm vein verification system / _cSoh Shi Chuan
264		1	_aKuantan, Pahang : _bUMP, _c2018
264		4	_c© 2018
300			_axiii, 93 pages : _billustrations ; _c30 cm. + _e1 CD ROM
336			_atext _2rdacontent
336			_atext _2rdacontent
337			_aunmediated _2rdamedia
337			_acomputer _2rdamedia
338			_avolume _2rdacarrier
338			_acomputer disc _2rdacarrier
347			_atext file _bPDF _2rda
500			_aFaculty of Electrical and Electronics Engineering
502			_aThesis (Master of Science) -- Universiti Malaysia Pahang – 2018
504			_aIncludes bibliographical references
520	3		_aBiometrics recognition system are getting more attention in efforts to protect our security and information in this world of digital impersonation. Palm vein recognition are well-known in biometrics recognition where it shows a high level of authentication. However, there is still an unsolved issued in accuracy due to the complexity and uniqueness of palm vein pattern. Low quality image provides unclear and low contrast image affecting the process although palm vein feature extraction is perfect. There were studies to investigate the possibility that fusion methods would improve or enhance the accuracy to a higher level. Image fusion is a method to collect necessary information from all input image with different sources and create an output image that ideally has information from input image. Fused image can provide more information than single input image that improve quality and applicability of data. In this work, image fusion algorithms based on Discrete Cosine Transform (DCT) in palm vein recognition is proposed. Input image will be divided into consecutive blocks and transformed into DCT coefficients. Fusion rule will be applied within the DCT coefficients and transformed back into fused image using inverse DCT. In this work, CASIA database is used to provide three types of wavelength spectrum which are 700 nm, 850 nm, and 940nm. There are four combination of image fusion that can be formed, dual combination with 700 nm and 850nm, 700 nm and 940 nm, 850 nm and 940 nm and triple combination of all wavelength. Multi-resolution DCT (MRDCT), Frequency Partition DCT (FPDCT) and Laplacian Pyramid DCT (LPDCT) image fusion is introduced on fusing more informative information from different types of wavelength and resulting in an image with finer details of vein patterns in the output image. In this work, triple combination of image fusion achieve better than dual combination of image fusion. By fusing three wavelength spectrums, MRDCT performed the best at 5.53% in EER rate compared to FPDCT and LPDCT. The conventional method such as Multi-resolution Singular Value Decomposition (MSVD), wavelet transform and Energy of Laplacian (EOL), were only able to achieve EER rate of 6.58%, 6.83% and 8.64% respectively. In addition to that, MRDCT with triple wavelength spectrum fusion showed a significant drop in EER by 9% compared with single 700 nm image, 7% compared with single 850 nm image, and 6% compared with single 940 nm image. It proved that MRDCT image fusion is suitable for palm vein recognition. For feature extraction, two types of local invariant feature based method was investigated, Scale Invariant Feature Transform (SIFT) and Speeded Up Robust Feature (SURF). SIFT algorithm achieved a reduction in EER rate by 12% in 700 nm, 8% in 850 nm, 7% in 940 nm compared with the SURF algorithm. The result shows that SIFT algorithm achieved a better recognition rate and extract more information and matching pairs compared to SURF algorithm. In conclusion, MRDCT image fusion with SIFT feature extraction are suitable to use in contactless palm vein recognition system.
610	2	0	_aFaculty of Electrical and Electronics Engineering _xDissertations
650		0	_aUniversities and colleges _xDisertations
650		0	_aTheses
942			_2lcc _cTHESIS