Within a Bayesian framework, Brady proposed the adaptive texture approach for more accurate description and applied
this model in texture segmentation with a neighbourhood-based algorithm. In this paper, the efficiency of the texture
model in Brady's segmentation method is investigated. In the segmentation experiments of Brodatz texture mosaics and a
remote sensing image, the results show that the good segmentation performance mainly owes to the
neighbourhood-based algorithm, but not Brady's texture description model. Moreover, this probabilistic model is applied
in texture classification with a MAP method. To improve the correct classification rate of the image bank, a method
combining the best adaptive texture description of each class is proposed and obviously improves the rate from 91% to
95%.
Compared with unimodal wavelet packet subbands, multimodal subbands have strong texture discriminatory power. The
existence of mulitimodal subbands in dual-tree complex wavelet packet transform is proved. Similar to the multimodal
subbands in real wavelet packet transform, there are shift-modal subbands in complex transform to capture the
periodicities running through the texture images. Furthermore, the stability of multimodal subbands in real transform is
investigated through a classification experiment. It is concluded that, to the textures with small and very regular
periodicities, stable multimodal subbands can be obtained.
For optical iris recognition, the eigen-images correlation recognition method is modified. The 2-D separable approximations of wavelet packet bases are constructed with the help of the cascade algorithm. Expanding the scale of basis selection, mutli-mother multi-vanishing moment joint best bases are chosen from the basis set of 25 mother wavelets including the mothers constructed by the lifting scheme. Using the corresponding eigen-images generated and the post-processing method based on statistic features, optical experiment is implemented. The experimental result agrees with the simulation result.
The eigen-image correlation recognition method proposed for optical image recognition is utilized in this paper for optical iris recognition. The best basis selection of this method is modified by introducing new criterion for evaluating wavelet packet bases. Using the cascade algorithm, the 2-D approximations of wavelet packet basis functions are computed. And the direct transforms of these wavelet packet bases make the larger scale of basis selection possible. From the wavelet packet basis set of 25 mothers, multi-mother multi-vanishing moment joint best bases are chosen for iris recognition. Their eigen-images are generated for optical experiments. Based on these best bases, an optical wavelet packet gray scale filter is designed and fabricated to improve the optical recognition. In a volume holographic correlation system, optical experiment results show this filter is efficient to achieve higher identification rate.
Iris, one important biometric feature, has unique advantages: it has complex texture and is almost unchanged for the lifespan. So iris recognition has been widely studied for intelligent personal identification. Most of researchers use wavelets as iris feature extractor. And their systems obtain high accuracy. But wavelet transform is time consuming, so the problem is to enhance the useful information but still keep high processing speed. This is the reason we propose an opto-electronic system for iris recognition because of high parallelism of optics. In this system, we use eigen-images generated corresponding to optimally chosen wavelet packets to compress the iris image bank. After optical correlation between eigen-images and input, the statistic features are extracted. Simulation shows that wavelet packets preprocessing of the input images results in higher identification rate. And this preprocessing can be fulfilled by optical wavelet packet transform (OWPT), a new optical transform introduced by us. To generate the approximations of 2-D wavelet packet basis functions for implementing OWPT, mother wavelet, which has scaling functions, is utilized. Using the cascade algorithm and 2-D separable wavelet transform scheme, an optical wavelet packet filter is constructed based on the selected best bases. Inserting this filter makes the recognition performance better.
Based on the cascade algorithm and the theory of 2-D separable wavelet transform, 2-D approximations of scaling and wavelet basis functions are computed and used in optical wavelet transform. The optical transform using these separable wavelet bases can be called optical separable wavelet transform. And the selection of mother wavelets is extended. Unlike 2-D discrete separable wavelet transform, optical separable wavelet transform does not have limitation on direction selectivity. Linearly combining multiple directional channels as a superposition filter, the transform of these bases can be fulfilled simultaneously and the transform results can be synthesized on the output plane. In this paper, 2-D scaling and wavelet basis functions of a biorthogonal wavelet, bior2.6, are calculated. Four directional channels combine into an oriented optical filter to increase the extracted feature energy in high frequency band. And simulation results are presented.
Although Optical wavelet transform has some advantages over discrete wavelet transform, but the mother wavelets to used are very few. That limits the signal processing ability of optical wavelet transform. Without scaling functions, the multiresolution analysis of a mother wavelet is not complete. In this paper, almost all the mother wavelets used in discrete wavelet transform are introduced into optical wavelet transform. Based on the analysis, we find whether the mother wavelets have analytical forms is not a necessary condition for implementing them in optical wavelet transform. Optical wavelet transform only needs to obtain the 2D approximations of wavelet functions. Then, with the cascade algorithm, the 1D approximations of scaling and wavelet functions are computed. By the scheme of 2D separable wavelet transform, the approximations of 2D scaling and wavelet functions are constructed. So mother wavelets frequently utilized in discrete wavelet transform are introduced into optical wavelet transform. With the increase of mother wavelet for selection, it is natural to classify optical wavelet transform into separable and non-separable cases as it does in discrete wavelet transform. Since the mothers introduced by the method in this paper are separable, they are included in the separable optical wavelet transform. And the advantages of the separable mothers are listed with corresponding examples.
Using iris feature, iris recognition attracts a lot of attention as a new and efficient personal identification technique in recent years. Compared with the frequently used methods of Daugman, Boles, et al., the dual multi-channel iris recognition system based on statistic features proposed by Yong Zhu, et al., has a unique and efficient algorithm. The algorithm processes gray iris image which is suitable to an Asian and takes good use of 2-D wavelet transformed irises. Moreover, they use statistic features to represent iris patterns which make their system more robust to errors caused in the image capturing stage. The recognition performance is better than the system of Wildes and approximates the system proposed by Daugman. But this system still has some open questions, such as, how wavelet filter channels influences the recognition and how to select wavelet channels. In this paper, we try to answer these questions. Via our analysis, it is proved that wavelet feature extraction can improve the identification rate and more wavelet filter channels results in better recognition. We also investigate the rule to choose the wavelet channels and conclude that high frequency channels are better than low frequency ones. Using this rule, we introduce wavelet packet channels to offer more useful information. The efficiency of this modification is shown by the experimental results.
The biometric feature, iris, has advantages in person identification, such as complex texture, almost unchanged throughout the lifespan. Compared with the famous methods propose by Daugman and Boles, the system of Yong Zhu, et al., not only takes good use of the 2D texture, but also is more robust for using statistic values of the wavelet transformed images as features for recognition. Because wavelet transform is time consuming, a volume holography opto-electronic hybrid system with high parallelism is constructed in this paper. Li Ding, et al., introduced wavelet packet transform into an optical recognition system based on volume holography to reduce the number of images stored in the photo-refractive crystal. By joint best basis selection, eigen-images corresponding to the best wavelet packet bases are generated and stored to replace the reference images. This replacement results in high compression. Theoretical analysis and experimental results both show their scheme achieves significant compression and accurate recognition at the same time. Wavelet packet compression is also utilized in our system. But the best basis selection algorithm is modified. For iris identification, we use the recognition capacity of each wavelet packet basis instead of the entropy because the latter is not for recognition. Furthermore, in the post-processing stage, we use statistic features, like Yong Zhu, to represent each iris pattern which makes the system more robust to the errors caused by optical system. So our system combines the advantages of optics parallelism, high image compression and accuracy of digital processing. Simulation results show a high identification rate is obtained.
Wavelet packet transform analyzes signals more finely than wavelet transform does. This advantage can be utilized in optical wavelet transform. To introduce wavelet packet transform into optics, mother wavelets that have scaling functions must be used. If the scaling function does not have analytical formula, its approximation can be computed using the cascade algorithm. With the refinement relationship, its wavelet function can by calculated. After the 1-D wavelet packet bases are obtained, 2-D separable wavelet packet bases can be constructed for optical wavelet packet transform. As an example, a volume holographic opto-electronic system is proposed to fulfill joint best basis selection for a face image bank with the mother Db3.
Adaptive wavelet transform, using an adaptive wavelet as a linear combination of different wavelets, has been applied in optical information processing successfully. Most of the adaptive wavelets are built by daughter wavelets from only one mother wavelet. In this paper, we construct a new adaptive wavelet for feature extraction under noise as prepossessing in face pattern recognition. To have the ability of de-noising, daughters for construction are generated by two different mother wavelets. We call this transform multi-mother adaptive wavelet transform. It is important that the new wavelet combines advantages of different mother wavelets. With artificial neural network, the parameters are adaptively computed. Simulation results show the wavelet is not only robust to noise, but also keep good recognition performance. In frequency domain, the spectrum of our wavelet is real and easy to be realized as an optical filter.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.