Zhijun Fang

A Fingerprint Recognition Scheme Based on Assembling Invariant Moments for Cloud Computing Communications

In cloud computing communications, information security entails the protection of information elements (e.g., multimedia data), only authorized users are allowed to access the available contents. Fingerprint recognition is one of the popular and effective approaches for priori authorizing the users and protecting the information elements during the communications. However, traditional fingerprint recognition approaches have demerits of easy losing rich information and poor performances due to the complex inputs, such as image rotation, incomplete input image, poor quality image enrollment, and so on. In order to overcome these shortcomings, in this paper, a new fingerprint recognition scheme based on a set of assembled invariant moment (geometric moment and Zernike moment) features to ensure the secure communications is proposed. And the proposed scheme is also based on an effective preprocessing, the extraction of local and global features and a powerful classification tool, thus it is able to handle the various input conditions encountered in the cloud computing communication. The experimental results show that the proposed method has a higher matching accuracy comparing with traditional or individual feature based methods on public databases.

Fingerprint matching based on extreme learning machine

Considering fingerprint matching as a classification problem, the extreme learning machine (ELM) is a powerful classifier for assigning inputs to their corresponding classes, which offers better generalization performance, much faster learning speed, and minimal human intervention, and is therefore able to overcome the disadvantages of other gradient-based, standard optimization-based, and least squares-based learning techniques, such as high computational complexity, difficult parameter tuning, and so on. This paper proposes a novel fingerprint recognition system by first applying the ELM and Regularized ELM (R-ELM) to fingerprint matching to overcome the demerits of traditional learning methods. The proposed method includes the following steps: effective preprocessing, extraction of invariant moment features, and PCA for feature selection. Finally, ELM and R-ELM are used for fingerprint matching. Experimental results show that the proposed methods have a higher matching accuracy and are less time-consuming; thus, they are suitable for real-time processing. Other comparative studies involving traditional methods also show that the proposed methods with ELM and R-ELM outperform the traditional ones.

Interpolation-Based Direction-Adaptive Lifting DWT and Modified SPIHT for Image Compression in Multimedia Communications

Image compression, as one of the key-enabling technologies in multimedia communications, has been paid much attention in the past decades, where the two key techniques discrete wavelet transform (DWT) and set-partitioning in hierarchical trees (SPIHT) have great influence on its final performance. Due to the proprieties of fast computation, low memory requirement, DWT has been adopted as a new technical standard for still image compression. But it did not make much use of the region information. Although several improved methods have been proposed that adopt direction-adaptive wavelet for using the geometric and spatial information, they still did not consider the texture information. Furthermore, the traditional SPIHT algorithm has the drawbacks of long bits output and time consuming. In this paper, we first propose a new technique named interpolation-based direction-adaptive lifting DWT. It can adaptively choose the best lifting direction and use the Lagrange interpolation technique to make prediction according to its local characteristics. This method makes good use of the image texture features. Then a modified SPIHT coding algorithm is presented. It improves the scanning process and can effectively reduce the coding bits length and running time. Experimental results demonstrate that our method can yield better results than traditional techniques.

Wavelet based approach for fusing computed tomography and magnetic resonance images

The multimodality medical image fusion plays an important role in clinical applications which can support more accurate information for physicians to diagnose diseases. In this paper, a new fusion scheme for computed tomography and magnetic resonance images based on wavelet analysis is proposed. After the images are decomposed by wavelet transform, the low frequency coefficients are performed with the maximal absolute values followed by verifying their consistency, and the high frequency coefficients are selected by a maximal local variance rule. The resultant image is then reconstructed by using the inverse wavelet transform with the combined wavelet coefficients. The performance of our method is qualitatively and quantitatively compared with some existing fusion approaches. Experimental results show that the proposed method can preserve more useful information and with higher spatial resolution.

Comparing studies of learning methods for human face gender recognition

Recently, some machine learning algorithms such as Back Propagation (BP) neural network, Support Vector Machine (SVM) and other algorithms are proposed and proven to be useful for human face gender recognition. However, they have lots of shortcomings, such as, requiring setting a large number of training parameters, difficultly choosing the appropriate parameters, and much time consuming for training. In this paper, we proposes a new learning method to use Extreme Learning Machine (ELM) for face gender recognition and compare it with other two main state-of-the-art learning methods for face gender recognition by using BP, SVM respectively. Experimental results on public databases show that ELM plays the best performances for human face gender recognition with higher recognition rate and faster speed. Compared with SVM, the learning speed of ELM is obvious reduced. And compared with BP neural network, it has faster speed, higher precision, and better generalization ability.

Multiwavelet video coding based on DCT time domain filtering

To improve the video encoding efficiency and deal with the real-time demerits of the multiwavelet time-domain filtering in the 3D multiwavelet, a multiwavelet video coding scheme based on DCT(Digital Cosine Transform) time-domain filtering is proposed in this paper. Firstly, the multiwavelet transformation is applied to get the spatial domain of the frame in video. Then, 3D transformation coefficients are obtained by the DCT time-domain filtering on the multiwavelet coefficients. Finally the 3D transformation coefficients are coded to be embed bitstream by using the symmetric 3-D SPIHT(Set partitioning in hierarchical trees). The experimental results indicate that the new method outperforms those based on 3D DMWT(discrete multiwavelet transform) in terms of PSNR(peak signal to noise ration) with better video quality at the same compression ratio, and it shows lower complexity.