Applied Mechanics and Materials Vols. 738-739

Abstract: In telemedicine, medical images are always considered very important telemedicine diagnostic evidences. High transmission delay in a bandwidth limited network becomes an intractable problem because of its large size. It’s important to achieve a quality balance between Region of Interest (ROI) and Background Region (BR) when ROI-based image encoding is being used. In this paper, a research made on balancing method of LS-SVM based ROI/BR PSNR prediction model to optimize the ROI encoding shows it’s much better than conventional methods but with very high computational complexity. We propose a new method using extreme learning machine (ELM) with lower computational complexity to improve encoding efficiency compared to LS-SVM based model. Besides, it also achieves the same effect of balancing ROI and BR.

Abstract: In this paper, a new approach to restoration of blurring image with content aware and texture enhancement is presented, in which the content aware term is embedded into the objective function of optimal decision. In this way, the content aware term can play a role to reinforce the restoration effect in the optimal decision. Thus the conventional total variation methodology is employed as a basic tool to solve such optimal decision problem with content aware in its objective function. Meanwhile, the image texture is enhanced by layer decomposition and edge preserving filter. A series of experiment results manifest the proposed algorithm can achieve better restoration effect than the conventional total variation method.

Abstract: Camera gesture information is important for making virtual 3D face model synchronize to the real face at view point, in other areas such as: human-computer interface, 3D object estimation, automatic camera control etc. Camera gesture is also very important. In this paper, we will introduce an algorithm to detect camera gesture from a single face image, through our algorithm, camera rotation angles in both x-axis and y-axis and distance can be gained to confirm camera gesture. Then we can set the camera gesture information to view point of virtual 3D face model. This can make can make virtual 3D face model synchronize to the real face at view point. Experiment results demonstrate the validity of our algorithm and the correct estimation rate is accredited for applying it into practice.

Abstract: Digital Elevation Model (DEM) is an important data source for topographic analysis, 3D visualization and satellite image ortho-rectification. This paper focused on the DEM extraction and accuracy assessment from ZY-3 satellite with 3 stereo images. DEM was extracted using three different stereo pair image groups composed of forward and nadir view images, nadir and backward view images as well as forward and backward view images. The accuracy of the DEM was indicated by root-mean-square error (RMSE) values. The results showed that the stereo pair of nadir and forward view images achieved the best accuracy, while the pair of forward and backward view images obtained the worst. This might be useful for the selection of the stereo pair images for extracting DEM using ZY-3 satellite images.

Abstract: In this article, a fast approach of video stabilization is presented based on haerachical estimmation of global motion and multi-thread parallel computation. At first, the structure-texture decomposition is adopted to tackle the negative effect from illumination changes. Then according to pyramid structure of image texture a hierachical model is built to estimate global motion parameters with least square method. Afterwards Gaussian smoothing method is employed to eliminate error accumulation in motion compensation. Meanwhile the multi-thread parallel computation is used to speed up the processing efficiency so as to achieve a real time performance for videos with 25fps.

Abstract: A gait recognition method based on wavelet packet decomposition (WPD) and Locality preserving projections (LPP) is proposed in this paper. The method includes the following steps, pretreatment, feature extraction by WPD and dimensionality reduction by LPP and classification of the test samples to a corresponding class according to the nearest neighbor classifier. The experiment results on the public gait database show the effectiveness of the proposed method.

Abstract: In this paper, we present an algorithm which detects human hand by skin color information in YCbCr and HIS color model. And for confirming special human hand we use circle rate of region to detect hand region because human hand have complex edge than other region, thus circle rate of hand region is usually more greater. For the recognition of detected hand, we use the Hausdorff to tracking the hand region. And we employed a recognition method based on PCA algorithm to recognize the hand gestures. The experimental results show that an algorithm plays an efficient effort for hand gesture recognition.

Abstract: In this paper, we mainly study on the relatively small moving target and its use in the badminton video, and we also study the relatively mining methods open space across the network sparring sport badminton video. Accurate extraction of a moving object based on the mining model is using to modified geometric features for the basic motion feature mining method to be analyzed, we than obtain video structural information based on statistics of people habits using the improved tracking algorithm, the results show that the method can solve the problem of missing and made a badminton video teaching system design.

Abstract: The license plate recognition system is an important part of modern traffic management. application which is very extensive. In this paper, a method to achieve three main modules split from the image pre-processing, license plate location and character. Image pre-processing module of this article is to image gray and step by Roberts operator edge detection. License plate positioning and segmentation using mathematical morphology is used to determine the license plate location method, and then use the license plate color information of color segmentation method to complete the license plate parts division. This article is to research its main part and use MATLAB to do the image processing simulation.

Abstract: In this paper, a matrix based feature extraction and measurement method, i.e.: multi-column principle component analysis (MCPCA) is used to directly and effectively extract features from the matrix. We analyze the advantages of MCPCA over the conventional principal component analysis (PCA) and two-dimensional PCA (2DPCA), and we have successfully applied it into face image recognition. Extensive face recognition experiments illustrate that the proposed method obtains high accuracy, and it is more robust than previous conventional face recognition methods.