Research Of Digital Watermarking And Video Object Tracking Method

Under recently, research on image processing has been widespread concerned globally. At this work, research on the digital watermarking for image and object tracking for video have been carried out respectively. These works focused on the new algorithm establishing and their feasibility demonstrating and capability testing. Mainly work contents and obtained results shown as following.(1)An algorithm based on second generation wavelet technology for embedding and extracting watermark into static image has been proposed. In this algorithm, the embedding and extracting of watermark in two-dimensional image is realized by lifted scheme wavelet transform. In other words, the host image is multi-resolution decomposed by wavelet transform. The characteristics of image are considered as masking parameter factor. Watermark is embedded by second generation wavelet transform after the low-frequency signal was approximated. In addition, watermark is repeatedly embedded into the middle band of image by threshold. The experimental results showed that the embedding and extracting of watermark in image could be realized effectively based proposed algorithm. Furthermore, this algorithm has a certain stability and resistance to attacking.(2) A fast human face tracking algorithm is proposed based on FloatBoost detection and exponential smoothing motion estimation. In this algorithm, weaker features are removed by FloatBoost, so it has a faster processing speed. At the same time, because exponential smoothing don’t need store a lot of historical data but considered the importance of each data to use all the historical data, it’s real-time. Experimental results showed that the accuracy of image processing by this algorithm is decided by the complexity of environmental. The simpler the better. When one to two faces are tracked as targets in system, it has steady-state of efficiency and good real-time. (3) A new measurement system for regional features matching has been proposed based on the idea of Chebyshev distance. Then, a new object tracking algorithm is proposed based this measurement system and the hypothesis testing for matching results. No matter motion patterns known or unknown, discuss and experiments are carried out. The results showed that object can be tracked effectively based on this algorithm, and the tracking precision higher than Kalman Filtering. The proposed algorithm has certain robustness.(4) Inspired by the statistical meteorology, a measurement named analogue deviation which can simultaneously reflect the similar of value and shape of data samples is established. Then, a normalized index system is proposed to describe the approximation of images information based on this measurement system. Of course, the index system establishing process meets mathematical logic and physical meaning. Based on SVM detection and analogue deviation, a new algorithm is proposed for human face tracking. Experimental results showed that the most similar sample for template can be found by matching based on analogue deviation. And it’s not susceptible to interference. It obtains good tracking accuracy in static images and dynamic videos.(5) An object tracking algorithm is proposed based on Harris corner detection and KNN restricting. In this algorithm, KNN classification is used to restrict the corner extracting by Harris algorithm. So redundant point can be removed, and simple but effective matching model of object characteristics can be established. Experimental results showed that it can effectively streamline corner, thereby reduce the computation of the matching process to improve the speed. Its also has a certain accuracy in video tracking by corner model matching after KNN restricting.