Font Size: a A A

PolSAR Image Compression Based On Statistical Characteristic And Dictionary Learning

Posted on:2016-02-05Degree:MasterType:Thesis
Country:ChinaCandidate:Y WeiFull Text:PDF
GTID:2348330488957208Subject:Engineering
Abstract/Summary:PDF Full Text Request
Synthetic Aperture Radar(SAR), as a tool in microwave remote sensing for high resolution, with the all-weather and all-time characteristics, are widely used in the remote sensing field, such as target identification, military reconnaissance and topography etc.. The Polarization SAR(PolSAR) which increases the polarization property is a further complement to SAR information, and greatly expanded its range of applications. However, due to PolSAR and SAR both have a huge amount of data, especially for PolSAR, which has caused tremendous pressure for the work of data transmission, thus it is very necessary for its data compression.In this paper, we do a detailed analysis to the commonly used two methods for SAR image compression, namely multi-scale decomposition and sparse representation with dictionary learning, then introduce the concrete realization of this two methods. Because of the difference from the usualSAR image, PolSAR image can not be compressed simply based on SAR image compression method, and the effective compression method for PolSAR image must consider the characteristics of itself.In this paper, we presents a PolSAR image compression method based on image similarity. We use the feature that PolSAR has a broad range of coverage area and it makes the corresponding image portion having similar characteristics. Combined with the sparse representation and dictionary learning method, we can not only effectively remove spatial redundancy, inter-channel redundancy but the eliminating redundant between image similarity blocks, and finally achieve better results on PolSAR image compression.This paper also presents a multi-directional RLS-DLA based PolSAR image compression method. Firstly, the method decomposes the image for multi-scale and multi-direction. Then, use RLS-DLA dictionary in every direction to get the best match for each direction. Finally, we follow the sparse representation and suitable coding strategies to get effective compression for PolSAR image. Since the dictionary used in sparse representation is characterized by high frequency sub-band coefficient in each direction of the converted image, so the compression process to protect the image edge details and important information of profile.Finally, the proposed algorithm in this paper are compared with the classical compression algorithm for the real PolSAR image compression processing, experimental results show that the algorithm proposed have a better effect on image compression both in visual aspects and evaluation indicators.
Keywords/Search Tags:SAR, PolSAR, Multi-Scale Decomposition, Sparse Representation with Dictionary Learning, RLS-DLA
PDF Full Text Request
Related items