Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research on Methods of Transform Based on the High Spectral Oil Image Compression

Author XuChangJian
Tutor LiChunGeng
School Dalian Maritime University
Course Computer Science and Technology
Keywords Oil Spill hyperspectral image Embedded zerotree wavelet Region-of-interest Texture
CLC TP391.41
Type Master's thesis
Year 2011
Downloads 78
Quotes 4
Download Dissertation

Hyperspectral Remote Sensing image has a lot of ad vantages.for example,it has many bands,image and spectral together,high resolution,rich data quantity,contains the continuous spectrum from Thermal infrared、Short-wave infrared to Near-infrared and Visible bands.So that it develops quickly in the oil spill detection agriculture、minerals, etc a wide range of applications.But hyperspectral data amount due to its huge, in actual application field is sacrificed.This huge data quantity, for airborne and space-borne of hyperspectral speaking, due to the limited bandwidth and storage, resulting in real-time data transmission can.Because lossy compression will bring information lost,so it has great influence for Subsequent research association of hyperspectral;adopt lossless, compression is small, amount of data effect is not obvious.Therefore,to a large extent on the establishment of a compressed data, while protecting the region of interest in hyperspectral image important information,to improve the reconstructed image compression mechanism of the late application, is very significant.This paper mainly studies the oil spill image region-of-interest (ROI) of hyperspectral image compression technology, includes the following three aspects:First, according to the characteristics of the oil spill area, using the texture method, to detect the region of interest.Second, for background image, using embedded zerotree wavelets, can according to the actual needs of the code rate adjustment coefficient of background, lossy compression; Interested area are using lossless compression of the huffman algorithm. This will not only save the valid information, to facilitate the follow-up work; and reduces the amount of image data.Third, the high-spectral image bands also exist large redundancy, this paper uses feature extraction to reduce the number of bands to achieve the purpose of compression.In this paper, Dalian and Qingdao’s offshore airborne hyperspectral oil image experiments were carried out, it’s entirety to preserve the original hyperspectral image based on the oil spill area of information and achieve better compression performance. The proposed compression method, able to analyze the late stages of oil spill to provide valuable data quickly.

Related Dissertations
More Dissertations