Motion Estimation Of Dynamic Image Sequence

The thesis focus on the technology of motion estimation of the image sequences, which is known as the principal field of computer vision and image sequences analysis. It is widely applied in the areas of national development and construction, economy and national defenses. As a result, the technology is usually task-launched and depended on different conditions. Consequently, based on the widely research of various estimation technologies, the thesis sorts them into three new categories as Intensity-based Matching techniques, Intensity-based differential techniques and frequency-based energy techniques. The thesis also focuses on the application techniques of the latter ones, which includes the strong and weak points according to different applications, robust and efficiency.Having projected the realistic world in to an image sequences model, we present the basic assumption of Intensity-based differential techniques: 'Constant assumption 'and derive the optical differential equation for the motion estimation. Then, we achieve H-S model for the motion estimation. Furthermore, the singularity of the optical differential equation is discussed and four groups of extra assumption are added to the equation to solve the problem.On frequency-based energy techniques, we proved the relation between the velocity and the energy of the moving pattern in frequency domain. Based on the link, we use the Gabor filter as the motion-detected filter and give its attributes and character of frequency response. Then the specific solution of motion estimation is provided.On the other side, the thesis point out the significance between the implementation issue and the techniques. We discussed the performance of different testing sequences and the give the reasonable error estimate mechanism.