影像金字塔的生成、压缩与存储模型研究

发布时间：2018-03-21 15:09

本文选题：金字塔　切入点：纹理滤波　出处：《太原理工大学》2016年博士论文　论文类型：学位论文

【摘要】：《国家中长期科学和技术发展规划纲要(2006~2020年)》确定了16个重大专项,高分辨率对地观测系统位列其中,遥感空间信息云服务RSCS(Remote-sensing and Spatial Information Cloud Service)是它的一个主要应用。RSCS首先需要解决两类模型的表示问题,即:地球空间的金字塔表示模型和空间金字塔的计算机表示模型。本课题针对空间金字塔的计算机表示模型开展研究,涉及金字塔数据的生成模型、金字塔数据的压缩模型和金字塔数据的组织模型三个主要方面。本论文将金字塔数据的组织模型表示成概念模型,逻辑模型和物理模型三个层次,指出:地球空间的金字塔表示模型实质上是建立三层模型的概念模型。当前金字塔数据是按照一种二元静态结构组织的,即用一个带有路径的文件名将概念建模得到的剖分索引映射到计算机文件系统:空间概念的一个剖分索引用作文件名,其关联的影像数据存入该文件中。由于文件系统元数据采用文件名+扇区链表的二元静态结构,文件内的数据元素是非结构化的黑箱,不支持随机访问,无法更深层次地表示金字塔数据集合之间,或层与层之间的关系,致使在剖份文件数量与数据处理粒度之间取得平衡成为一个两难问题。本研究建立了空间金字塔的三元动态组织结构:它的概念模型是地球空间金字塔表示模型建立的剖分索引;本课题提出一种保持地址不变的金字塔存储结构PL-SS,作为它的物理模型;在概念模型和物理模型之间建立一种映射关系,作为它的逻辑模型;逻辑模型具体实现为一种多维元数据的新型文件系统,可以支持云存储的随机并行访问,是一种云存储文件系统CSFS。针对空间金字塔三元动态组织结构的定位体系,本论文指出:概念空间定位体系是经纬度坐标系,概念建模的任务是将地球空间的概念映射到经纬度坐标系;在物理空间包括像素坐标系和逻辑地址坐标系两个定位体系,物理建模的任务是将像素坐标系映射到存储坐标系;本论文一个主要工作内容是建立一个逻辑层将概念层映射到物理层,定义一种公共坐标系用作逻辑空间的定位体系。在概念空间和物理空间,公共坐标系分别建立了和像素坐标与逻辑地址坐标的一一对应关系,形成一种全新的金字塔计算机表示模型框架。由于金字塔数据是计算机生成的,各层没有原始数据,无法使用信噪比来评价金字塔的数据生成质量。本课题提出用重采样滤波器对偶RSFP来获得信噪比的一个近似,在此基础上筛选一种基于纹理搜索的降采样滤波方法TDFA,用于金字塔生成模型比此前最好滤波效果提高4.51~5.70 d B。基于重采样滤波器对偶RSFP,本课题提出了一种层间压缩方法ICRS,在重建图像PSNR高于40d B时,压缩率比AVS的I帧算法提高19.70倍,是目前一种效率最高的金字塔压缩模型。
[Abstract]:"National long-term science and technology development plan (2006~2020)" identified 16 major projects, high-resolution earth observation system in remote sensing, spatial information cloud services RSCS (Remote-sensing and Spatial Information Cloud Service) is one of the main problems with.RSCS should be said, first needs to solve two types of models of it namely: the earth's space representation model and spatial model of Pyramid Pyramid computer. The topic for the space of Pyramid computer representation research model, data generation model involving Pyramid, Pyramid and Pyramid data compression model data organization model in three aspects. The paper said Pyramid's data organization model into conceptual model. The logical model and physical model of three levels, pointed out: the earth's space in Pyramid said the model is essentially a three layer model concept Read the current data model. Pyramid is in accordance with a two yuan of static structure of the organization, to obtain the file name conceptual modeling with a path subdivision index is mapped to the computer file system: a section of the concept of space division index paper, the number of its associated image stored in the file. According to the two yuan of static structure of file system metadata based on file name + sector list of the data elements within the file is unstructured box, do not support random access, not deeper between Pyramid data set, or the relationship between layer and layer, resulting in a balance between the number of documents and data processing section size become a dilemma. This study established three yuan dynamic organization of space in Pyramid: its conceptual model is to establish the model of Pyramid said earth space subdivision index; this paper proposes a Pyramid keep the storage structure of PL-SS address, as its physical model; establishing a mapping relationship between conceptual model and physical model, logical model as its logical model; implementation of new file system as a multidimensional metadata, randomized parallel access can support cloud storage, cloud storage is a file system according to the CSFS. positioning system in Pyramid three yuan dynamic organization structure, this paper points out: the concept of space positioning system is the longitude and latitude coordinates, conceptual modeling is the task of concept mapping in earth space to latitude and longitude coordinates; including two positioning system coordinates and pixel coordinates in the logical address space of physics, physical modeling the task will be mapped to the storage coordinates of the pixel coordinate system; this paper is a main content of the work is to build a logical layer concept layer is mapped to the physical layer, the definition of a The positioning system of public coordinate system as a logical space. In the concept of space and physical space, a common coordinate system are established and the corresponding relationship between pixel coordinates and the coordinates of the logical address, the formation of a new Pyramid computer model framework. Because Pyramid is the data generated by the computer, not the original data of each layer, data quality cannot use the signal-to-noise ratio evaluation of Pyramid. This paper uses dual RSFP resampling filter to obtain an approximate signal-to-noise ratio, based on the screening of a TDFA filter method based on texture search reduction for Pyramid than the previous generation model, the best filtering effect to improve 4.51~5.70 D B. resampling filter based on dual RSFP in this paper, a layer of ICRS compression method, image reconstruction in PSNR is higher than 40d at B compression rate than I-frame algorithm AVS increased 19.70 times, is At present, the most efficient Pyramid compression model.

【学位授予单位】：太原理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP751

【相似文献】