SAR (Synthetic Aperture Radar) image segmentation method based on weighted gamma hybrid model integrated with spatial information

Abstract

The invention discloses an SAR (Synthetic Aperture Radar) image segmentation method based on a weighted gamma hybrid model integrated with spatial information, and belongs to the technical field of image processing. The method comprises steps of reading an SAR image to be segmented; establishing an SAR image statistical model by using the weighted gamma mixture model fused with the spatial information, and obtaining an SAR image segmentation model based on the weighted gamma mixture model fused with the spatial information according to the Bayesian theorem; setting the number of iterations, solving a shape parameter by using a Markov chain Monte Carlo method in each iteration process, and solving a component weight and a scale parameter by using an expectation maximization method so as to obtain an optimal solution of the SAR image segmentation model based on the weighted gamma hybrid model fused with the spatial information; and outputting an SAR image segmentation result. Pixel spectral information and spatial information are effectively utilized, the problems of difficult parameter solving, large calculation amount, low efficiency and the like caused by a complex shape parameter structure are solved, and the SAR image segmentation precision is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a SAR image segmentation method based on a weighted gamma mixture model incorporating spatial information. Background technique [0002] Since the statistical distribution of pixel spectral measures in SAR (Synthetic Aperture Radar) images approximately obeys a certain probability distribution, the SAR image segmentation method based on statistical models has received extensive attention and research from scholars. This method assumes that the pixel spectral measures obey With the same known probability distribution, the image segmentation problem is transformed into a model parameter solving problem through maximum likelihood estimation, and then image segmentation is realized. [0003] The finite mixture model is a statistical model that can effectively model the statistical distribution of pixel spectral measures. It is defined by the weighting of multiple components....

AI Technical Summary

Problems solved by technology

[0004] At present, the proposed SAR image segmentation method based on the gamma mixture model has the following problems: (1) The gamma mixture model is difficult to accurately establish the statistical model of SAR images

Due to the inherent speckle noise of the SAR image itself, the statistical distribution of the pixel spectral measurement in each region presents complex statistical characteristics, such as heavy tails, peaks, and double peaks, while the gamma distribution presents a single peak and a heavy tail on the right. Accurately fitting complex statistical distributions, and the gamma mixture model using gamma distribution as a component is difficult to meet the requirements of establishing a SAR image statistical model; (2) Statistical modeling of pixel spatial information increases the complexity of the segmentation model

In order to avoid the influence of speckle noise in the SAR image on the segmentation results, the SAR image segmentation method based on the gamma mixture model usually uses the attributes of the neighborhood pixels to establish the prior distribution of the component weights, and according to the Bayesian theorem, the spatial information of the image is introduced into Segmentation model, but at the same time increases the complexity of the parameter structure in the segmentation model, which brings difficulties and challenges to the design of subsequent parameter solving methods; (3) It is difficult to accurately solve the parameters of the segmentation model

Accurately solving the parameters of the segmentation model is a necessary condition for achieving high-precision image segmentation, but in the SAR image segmentation method based on the gamma mixture model, the shape parameters of the gamma distribution exist in the form of gamma functions, so the shape parameter structure comparison Complex, need to use parameter optimization method to solve the shape parameters, but this kind of method leads to a large amount of calculation and low efficiency

In addition, due to the consideration of local pixel space information in the prior distribution of component weights, the analytical formula of component weights cannot be deduced through maximum likelihood estimation, and parameter optimization methods are also required to solve component weights, resulting in a large amount of calculation for such methods ,low efficiency

[0005] In summary, due to the problems of inaccurate statistical modeling, complex segmentation model, and difficulty in solving parameters in the gamma mixture model, the accuracy of the SAR image segmentation method based on the gamma mixture model needs to be improved.

Images