Abstract We propose a new image registration scheme for remote sensing images. This scheme includes three steps in sequence. First, a segmentation process is performed on the input image pair. Then the boundaries of the segmented regions in two images are extracted and matched. These matched regions are called confidence regions. At last, a non-linear optimization is performed only in the matched regions to get a global set of transform parameters. Experiments show that this scheme is more robust and converges faster than registration of the original image pair. We also develop a new curvematching algorithm based on curvature scale space to facilitate the second step. Keywords image registration · curve matching · curvature scale space 1

This paper describes the need for and the proposed designing of the high performance quantitative retrieval model to be used on Computational Grid for study of aerosol properties, with particular emphasis on Aerosol Optical Thickness (AOT) determination. A methodology using multi-resource remotely sensed data and adapting available aerosol retrieval model in a Grid environment is demonstrated. The algorithm comprises two complementary parts, collectively used in a distributed application. This paper focused on parallelization method based on a resource management and task partition strategy. A module, called DPPA (Dynamic Partition Points Algorithm for workload estimation), is designed as a portable technology for developing and deploying Grid execution in a generic data parallel paradigm. Experimental results are presented in a realistic application, using data collected by MODIS over China land. Derived result and computing performance of the proposed algorithm is given using the Grid test-bed at the Institute of Remote Sensing Applications of Chinese Academy of Sciences (IRSA, CAS). Combined, the experimental results show that Gridenabled model allowed on-demand large volume of ground-based data assimilation with parameters, and achieved substantial reductions in computational times. The research gives a thoughtful perspective on the potential of applying high performance computing practices to remote sensing quantitative retrieving problems. 1.

A 3D brain digital atlas is an important tool for neuroscience research. Many different imaging technologies such as MRI, CT, PET, microscopic imaging, and wet section photography, etc., are available. Among the technologies, cryosectioning followed by wet section photography can yield the highest resolution image of a full brain section. However, the images in the raw image sequence are not aligned with each other and therefore must be spatially registered. This dissertation describes several new methods employed by the atlas construction process. The most important components of the image registration process are the objective function and the optimization strategy. Pairwise image registration is inappropriate due to the lack of consideration of global coherence. A novel objective function called minimum entropy of bad prediction (MEBP) is proposed. MEBP is based on information theory and can be used for multi-modal image registration as well as image sequence alignment (ISA). In ISA, MEBP concurrently take multiple images into consideration and therefore can yield better alignment result. The optimization algorithm is a new hybrid method composed of density-based clustering algorithm, multi-resolution method and simplex