| Video compression is a pervasive technology. It represents an image sequence with a string of binary data, and it removes the need for video-specific transmission and storage systems. Applications relying on video compression include digital satellite, digital cable, DVD, and digital video recorder products, as well as web streaming and video conferencing devices. Scientific, security and remote sensing applications benefit from compressed video as well.; This dissertation addresses the errors introduced during video compression. These errors arise when the original image sequence is complex, and they are exacerbated when only a small number of digital bits are available to represent the sequence. To mitigate coding errors, the original images are often smoothed and re-sampled before compression. This degrades the data as well, and it effectively trades one error type for another.; Addressing the combined errors of blurring, sampling and video compression is the major goal of this work. This provides a novel contribution to the field of signal processing, while at the same time contributing to the specialized areas of post-processing, super-resolution and deblurring. Each of these specializations focuses on one of the three error types. Major contributions of the dissertation include a post-processing algorithm that exploits the motion vectors, a super-resolution algorithm that jointly estimates the displacements and high-resolution information from the compressed data, and a deblurring method that incorporates the bit-stream quantizers. |
