Jiuqiang Han

Multimodal biometric authentication based on score level fusion using support vector machine

Fusion of multiple biometrics for human authentication performance improvement has received considerable attention. This paper presents a novel multimodal biometric authentication method integrating face and iris based on score level fusion. For score level fusion, support vector machine (SVM) based fusion rule is applied to combine two matching scores, respectively from Laplacianface based face verifier and phase information based iris verifier, to generate a single scalar score which is used to make the final decision. Experimental results show that the performance of the proposed method can bring obvious improvement comparing to the unimodal biometric identification methods and the previous fused face-iris methods.

Three-dimensional reconstruction of endoscope images by a fast shape from shading method

Three-dimensional (3D) reconstruction from images has been one of the fundamental problems in the domain of computer vision. Shape from shading (SFS) is an approach to obtain the shape of an object from a single intensity image. This paper presents a 3D reconstruction approach for endoscope images using a fast SFS method. A new image irradiance equation of endoscope images is established based on the facts that the image is formed under perspective camera projection and the surface is illuminated by a near point light source. The light source is assumed to be located at the optical center of the camera and the attenuation of the illumination is taken into account. Then the image irradiance equation is derived as a nonlinear partial differential equation (PDE), which is associated with a static Hamilton–Jacobi equation considering the boundary conditions. The viscosity solution of the resulting Hamilton–Jacobi equation is approximated by using a new iterative fast marching method. The proposed approach is evaluated on both synthetic and real medical endoscope images and the experimental results show that the proposed approach is fast and accurate.

A Novel Shape from Shading Algorithm for Non-Lambertian Surfaces

3D shape reconstruction is a classical problem in the field of computer vision. Shape from shading (SFS) is one approach to obtain the 3D shape from a single shading image. And Lambertian surfaces are an important assumption in traditional SFS methods. However, Lambertian reflectance model used to express the reflectance property of diffuse reflection is proved to be inaccurate. In this paper, we present a novel SFS algorithm for non-Lambertian diffuse surfaces. The Oren-Nayer reflectance model is used to approximate the reflectance property of the diffuse reflection. The image irradiance equation for diffuse surfaces is established with the assumptions that the camera performs an orthographic projection and its direction is as same as the point light source’s. Then, the equation is derived as an Eikonal partial differential equation (PDE) which consists of the shape of the surfaces. The viscosity solution of the resultant PDE is solved by using the high-order Godunov fast sweeping scheme. Experiments are conducted on both synthetic and real images and the results show the effectiveness of the presented algorithm.

Information fusion in personal biometric authentication based on the iris pattern

Information fusion in biometrics has received considerable attention. This paper focuses on the application of information fusion techniques in iris recognition. To improve the reliability and accuracy of personal identification based on the iris pattern, this paper proposes the schemes of multialgorithmic fusion and multiinstance fusion. Multialgorithmic fusion integrates the improved phase algorithm and the DCT-based algorithm, and multiinstance fusion combines information from the left iris and the right iris of an individual. Both multialgorithmic fusion and multiinstance fusion are carried out at the matching score level and the support vector machine (SVM)-based fusion rule is utilized to generate fused scores for final decision. The experimental results on the noisy iris database UBIRIS demonstrate that the proposed fusion schemes can perform better than the single recognition systems, and further prove that information fusion techniques are feasible and effective to improve the accuracy and robustness of iris recognition especially under noisy conditions.

Novel image fusion algorithm based on human perception

With the rapid development of numerous imaging sensors, a variety of image fusion algorithms have been proposed. However, these methods only transfer absolute information into the fused image and neglect the relative information contrast. Nevertheless, the contrast is the key stimulus variable represented by a neurons activity. We propose a new image fusion method inspired by the human visual system. The new method uses the contrast to represent the salient features of an image. First, we use the polyharmonic local sine transform to decompose the source image into two components: the polyharmonic component and the residual. Next, we compute the contrast of every pixel by dividing the residual by the average value of the polyharmonic component. The polyharmonic component and residual are fused separately with different fusion rules. We can easily obtain the fused image through directly adding the composite polyharmonic component and the composite residual. Compared with the existing image fusion methods, such as the Laplace pyramid, shift-invariant wavelet, and contourlet, the inverse transformation is easy. The results demonstrate that the proposed algorithm is effective and is superior to the conventional image fusion algorithm in terms of the mutual information.

3D shape reconstruction of medical images using a perspective shape-from-shading method

A 3D shape reconstruction approach for medical images using a shape-from-shading (SFS) method was proposed in this paper. A new reflectance map equation of medical images was analyzed with the assumption that the Lambertian reflectance surface was irradiated by a point light source located at the light center and the image was formed under perspective projection. The corresponding static Hamilton?Jacobi (H?J) equation of the reflectance map equation was established. So the shape-from-shading problem turned into solving the viscosity solution of the static H?J equation. Then with the conception of a viscosity vanishing approximation, the Lax?Friedrichs fast sweeping numerical method was used to compute the viscosity solution of the H?J equation and a new iterative SFS algorithm was gained. Finally, experiments on both synthetic images and real medical images were performed to illustrate the efficiency of the proposed SFS method.

Fast viscosity solutions for shape from shading under a more realistic imaging model

Shape from shading SFS has been a classical and impor- tant problem in the domain of computer vision. The goal of SFS is to reconstruct the 3-D shape of an object from its 2-D intensity image. To this end, an image irradiance equation describing the relation between the shape of a surface and its corresponding brightness variations is used. Then it is derived as an explicit partial differential equation PDE. Using the nonlinear programming principle, we propose a detailed solu- tion to Prados and Faugerass implicit scheme for approximating the viscosity solution of the resulting PDE. Furthermore, by combining im- plicit and semi-implicit schemes, a new approximation scheme is pre- sented. In order to accelerate the convergence speed, we adopt the Gauss-Seidel idea and alternating sweeping strategy to the approxima- tion schemes. Experimental results on both synthetic and real images are performed to demonstrate that the proposed methods are fast and accurate.

Approaches to realize high precision analog-to-digital converter based on wavelet neural network

A new method is proposed in this paper to implement the high precision analog-to-digital converter (ADC) with low precision ADC based on two-stage conversion. Because the main error of ADC is non-linear, an algorithm using wavelet neural network for compensating error and non-linearity of ADC is proposed, which has faster speed quality convergence and higher precision than BP neural network. By studying the theories and scope of ADC errors, the wavelet neural network is used to deal with the non-linearity part of ADC error, which simplifies the network structure and requires shorter training and less iterations of learning. The experimental results show that with the wavelet approximation, the non-linearity of ADC can be reduced markedly, and the conversion speed of ADC can maintain maximum.

Design optimization of DWT structure for implementation of image compression techniques

The optimal VLSI architecture and algorithm for computing lifting based wavelet is presented. For the (5, 3) filter, the division by four or by two is implemented with look-up table (LUT). But for the (9, 7) filter, the operation is more complex. We present the fragment look-up table algorithm implements the multipliers. The algorithm can reuse the small memory to realize high precision for the multipliers (scaling by K or 1/K). Experimental results demonstrated that the proposed approach is effective in reduce chip size and power consume. The algorithm is also simpler and faster, and can be implemented easily for SOC integration.