Shanzhen Lan

Learn More
Vehicle analysis is an important task in many intelligent applications, which involves vehicle-type classification(VTC), license-plate recognition(LPR) and vehicle make and model recognition(MMR). Among these tasks, MMR plays an important complementary role with respect to LPR. In this paper, we propose a novel framework to detect moving vehicle and MMR(More)
Facial expression details can largely enhance the realistic facial animation. In this paper, we proposed a method to realize real-time control of 3D facial subtle expression, such as wrinkles and bulges, based on a set of normal maps. At runtime, we used a kinect to get the animation units for calculating control parameter. Normal blending technology was(More)
In data-driven mesh-based P2P live streaming system, the playback buffer management affects the audiences' quality of experience (QoE) significantly. Due to the instability and heterogeneity of P2P networks, the buffer is not always ready for playback. It is not only associated with the media timestamp, but also with the data availability in its neighbors.(More)
HyperText Markup Language (HTML) is the main markup language for web pages. HTML5 is a new generation of HTML standards. With the development of HTML5, it has wide range of applications in the multimedia direction. This new standard first describes the Video and Audio tag, and then elements of Video and Audio are discussed, pointing out that the support of(More)
TV logo recognition plays an important role in video content understanding. So far, Real-time TV logo recognition under complex background based on single frame is still a very challenge task. By analyzing the characteristics of TV logo, a new TV logo recognition algorithm based on multiple feature fusion via hierarchical matching was proposed in this(More)
In order to improve the robustness of the existing character recognition systems and increase their recognition ratio under various bad illumination conditions, we proposed a fast image enhancement algorithm based on illumination-reflectance model, by using an approximate edge-preserved filter. Experimental results showed that our algorithm can greatly help(More)
With the diversification and ever-changing forms of video programs and advertising information, the application of video corner-logo is more and more widely used. Corner-logo detection can be used for video content recognition and interaction, information analysis, monitoring and other aspects of advertisements. An effective corner-logo detection algorithm(More)
Car model recognition plays an important role in the road auto tolling system and the traffic volume statistics. In this paper, we present an improvement on car model recognition based on convolutional neural network. Firstly, our car image database which includes train and test images has been established. Then we preprocess our car model images using(More)