Motion Estimation for Digital Video

摘要

The estimation of motion from a sequence of images is important for dynamic scene analysis, which includes robot vision and medical applications, as well as for digital video, which includes video coding, image (sequence) estimation and restoration, and video standard conversions involving de-interlacing and frame rate conversions. The thesis is concerned with the design of motion estimation algorithms with application in the broad field of digital video. In Chapter 2, the authors analyze the possible representations of motion and image data that can be adopted for motion estimation. In Chapter 3, the authors focus on the use of stochastic models that account for the spatial correlation between vectors in the motion field. In Chapter 4, the authors propose to extend the spatial smoothness models in order to incorporate the temporal correlation between vectors in consecutive motion fields. In Chapter 5, the authors propose an estimation approach that is based on spatial smoothness, but which takes into account occlusion effects due to overlapping, moving objects. In Chapter 6, the presence of noise in the observed images is explicitly taken into account. In Chapter 7, the authors discuss the application of high-definition motion fields in digital video coding.