Two new image processing algorithms and a framework for Internet-based image processing.

摘要

In this dissertation, we present two new image processing algorithms, a wavelet-based multiresolution statistical model for texture image, and an image segmentation approach using multigrid MRF optimization and perceptual considerations. The first algorithm leads to significant improvements in modeling highly complex images (e.g., ones with long-range and non-linear correlation structures). The second algorithm provides significant improvements in image segmentation when the image regions are characterized by non-stationary intensity or texture, as is the case in many real-world images.; To enable these two algorithms and other similar functionalities to be used for image processing research and education over the Internet, we have developed a new framework/software-environment for distributed image processing over the Internet. This integrated distributed framework contains two key features. First, it is highly interoperable. Software components developed by different people, using different programming languages and running on different platforms are wrapped into distributed objects that can be executed by a user anywhere over the Internet, without the user having to have the source code or do any installation or compilation. Secondly, it is highly integrated. Under a single client environment, a user can not only view static HTML documents but can also run interactive image processing programs, which may trigger location transparent streaming video clips (with audio). This framework is demonstrated in a prototype for remote education and software presentation in image processing. However, it can be easily extended to many other application areas where interoperability and resource sharing are desirable, such as education in digital signal processing, business, mathematics, physics, or other areas such as employee training and charged software consumption.