Zhang Jianqi

Face recognition method based on support vector machine and particle swarm optimization

Face recognition belongs to the problem of non-linear, which increases the difficulty of its recognition. Support vector machine (SVM) is a novel machine learning method, which can find global optimum solutions for problems with small training samples and non-linear, so support vector machine has a good application prospect in face recognition. In the study, the novel face recognition method based on support vector machine and particle swarm optimization (PSO-SVM) is presented. In PSO-SVM, PSO is used to simultaneously optimize the parameters of SVM. FERET human face database is adopted to study the face recognition performance of PSO-SVM, and the proposed method is compared with SVM, BPNN. The experimental indicates that PSO-SVM has higher face recognition accuracy than normal SVM, BPNN. Therefore, PSO-SVM is well chosen in face recognition.

Joint DWT and Integral Imaging for a Three-Dimensional Watermarking System

In recent years, three-dimensional digital watermarking has become a new hotspot in optical information security. This paper presents a new three-dimensional digital watermarking method based on integrated imaging. Firstly, three-dimensional digital watermarking is generated by computational integral imaging system that is implemented with smart pseudoscopic-to-orthoscopic conversion model. Secondly, discrete wavelet transform algorithm is applied to embed and extract the three-dimensional digital watermarking. Finally, three-dimensional digital watermarking is identified and showed by integral imaging system. The feasibility and effectiveness of the proposed method is demonstrated by experiment, the new method is not only able to meet the requirements of robustness and security, but image quality and display quality achieve these criterions of the human visual model. The proposed method opens up a new research perspective for the copyright protect of three-dimensional digital multimedia products.

Target-Acquisition-Performance-Model-Based Quality Metric for Multisensor Image Fusion

For applications concerning target detection and tracking, an image fusion quality metric that considers only the information preservation from the source images is not enough. A new quality metric for image fusion that combines the structural information preservation from the source images and the influence of the fusion process on the detection probability is proposed. Thus, the results of the proposed measure are more consistent with subject evaluations.