Research On Generation, Compression, And Storage Model Of The Image Pyramid

In the country to develop a “The National Medium- and Long-Term Program for Science and Technology Development(2006-2020)”, high resolution earth observation system is listed at one of its 16 major special projects, the remote-sensing and spatial information cloud service(RSCS) is one of its main applications. Said the problem of RSCS, it is first needs to solve two types of representation model namely: the spatial Pyramid representation model of the earth, and the computer representation model of the spatial Pyramid. This thesis researches on the computer representation model of the spatial Pyramid, involving the three main aspects: the Pyramid data organization model, the Pyramid data generation model and the Pyramid data compression model. In essence, the Pyramid data organization model is a ternary architecture of a conceptual model, a logical model and a physical model, and its conceptual model is modeled using the Geospace pyramid representation model.Current the Pyramid data organization structure belongs to a kind of two-dimensional static model, namely: the partition index obtained by concept modeling, and its relevant data are tightly combined into a who le of file, the data elements in this file internal is an unstructured blackbox, no supports random access, unable to deeply express the relationship between the Pyramid data sets, or between its layers. As a result, the balance between the number of the split file and the granularity of the data processing becomes a dilemma problem. It is the blame on two-dimensional static metadata structure of a file name and a sector’s chain tab le of the file system.This study established the spatial pyramid ternary dynamic organization structure: its conceptual model is the partition index coming from the Earth spatial Pyramid representation model; this study proposes a pyramid storage structure PL-SS to keep its address no changing and is used as its physical model; establish a mapping between the conceptual model and the physical model, serving as its logical model; to concretely implement a logical model, this thesis invented a novel file system owning multi-dimensional metadata to support parallel random access of the cloud storage, so it is a cloud storage file system CSFS.For the positioning system of space pyramid ternary dynamic organizational structure, this paper points out: the system of localizating concepts space is the latitude-longitude coordinate system, conceptual modeling task is to map a concept of geo-spatial to a position of the latitude-longitude coordinate system; in phys ical space, including p ixel coordinate system and logical addresses coordinate system two positioning system, physical modeling task is to take the p ixel coordinate system mapping onto store coordinate system. One of the main content of this paper is to create a logical space, as to mapping the concept space onto the physical space, a common coordinate system is defined as the logical space positioning system. For the conceptual layer and the physical layer, this common coordinate system establishes the correspondence relation between pixel coordinates and the the logical address, forming a new framework of computer representation model of the pyramid.Because pyramid data is computer generated, its each layer without origina l data cannot use signal-to-noise ratios to evaluate pyramid data quality. In this paper proposed a resampling filter pair RSFP to obtain an approximation of the signal to noise ratio, on the basis of, selected out a down-sampling filtering approach TDFA based on searching image texture, for Pyramid generation model to enhance PSNR 4.51~5.70 d B than the traditional best result. Based on the resamp ling filter pair RSFP, this thesis proposed an method of interlayer compression pyramid ICRS, guaranteeing over 40 d B of reconstructed image quality, its compression ratio 19.70 was higher than that of AVS I-frame coding arithmatic, which is one of the most efficient the pyramid compression model.