Since the traditional K-means algorithm can hardly predefine the number of clusters, and performs sensitively to the initial clustering centers and outliers, which may result in unstable and inaccurate results, an improved density-based K-means algorithm was proposed. Firstly, high-density trajectory data points were selected as the initial clustering centers to perform K-means clustering by considering the density of the trajectory data distribution and increasing the weight of the density of important points. Secondly, the clustering results were evaluated by the Between-Within Proportion (BWP) index of cluster validity function. Finally, the optimal number of clusters and clustering were determined according to the clustering results evaluation. Theoretical researches and experimental results show that the improved algorithm can be better at extracting the trajectory key points and keeping the key path information. The accuracy of clustering results was 28 percentage points higher than that of the traditional K-means algorithm and 17 percentage points higher than that of the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The proposed algorithm has a better stability and a higher accuracy in trajectory data clustering.

A new panorama generation algorithm for mobile devices was proposed to solve the problem of low stitching speed, more memory consumption, chromatic aberration and ghosting. First, the color correction was performed on source image sequences to balance color and luminance between adjacent images. Then ghosting artifacts were detected when stitching panorama. If a ghosting artifact was found, the corresponding object in the source image would be located, and a gradient domain object removing and region filling operation would be applied to remove the moving object. In addition, Poisson blending was used to further smoothen color transitions and hide visible seams. The time of Poisson blending was greatly reduced after color correction, and a unique memory allocation mechanism was also applied during image stitching process to decrease memory consumption. Finally, the method was tested on a mobile phone with configuration of 332MHz processor and 128MB memory by taking photos of resolution of 1280×720 under different illumination conditions, and compared with the traditional global panorama stitching algorithm by stitching 2 to 9 original sequential images, the memory consumption of global panorama stitching algorithm was from 12.3MB to 23.6MB, while the proposed method took up less memory, only from 9.9MB to 14.5MB. The experimental results show that this method eliminates image seams and "ghost" effect more thoroughly with high mosaic speed and low memory consumption, and the quality of generated panoramic images is better, thus it can be used on mobile devices for panoramic image generation.

Image processing technology depends on the quality of the visual saliency map to obtain better results. The existing visual saliency detection method usually can only detect and get rough visual saliency attribute graph, seriously affecting the image processing results. This paper put forward a method of using Bayes theorem and statistical learning of visual saliency detection to detect the visual saliency property of image. The method was based on Bayes theorem of static image top-down significant and overall significance, and combined the top-down knowledge and the down-top significance. For the synthetic integration of characteristics, all the factors related to the weight parameter were studied by using linear model with the linear weighting combination method and regularized neural network combined with nonlinear weighting method. The ROC curves of the bottom-up visual saliency model in two fixed data set for quantitative evaluation, show that the effect of non-linear combination is better than that of linear combination.

 The secure problems of E-cheque used in electronic commerce payment were studied. Based on the analysis of the developing circumstance and the security gap of it, a security guarantee system combining watermarking and digital signature were proposed. The double entity authentication and watermarking content authentication made it impossible to gain illegal access to E-cheque, to edit or to forge it. According to the analysis, the security, creditability and authenticity of the E-cheque can be achieved by the system.