Research On The Remote Sensing Image Segmentation Algorithm Based On Wavelet Transform

Image segmentation is a very important image analysis techniques,generally subject to a wide range of research and attention in recent years.Ithas thousands of image segmentation methods, and a large number ofapplications in the actual engineering.Image segmentation is the basis ofthe analysis of the image processing, and the key steps of the targetexpression, characteristic measurement, occupies an important role in theimage project.The wavelet transform is a time-frequency analysis tool, caneffectively extract information from the image by decomposing the imageinto different scales for processing.It has good local characteristics in thetime-frequency domain, overcomes the disadvantages of the analysis ofnon-stationary signal of the Fourier transform and add window Fouriertransform.Firstly,this paper makes a brief overview of basic imagesegmentation method and introduces the traditional method of imagesegmentation.For complex remote sensing images, combined with theadvantages of wavelet transform in image processing which focuses on thetheory of wavelet transform, and then eliminate the noise of remote sensingimages using wavelet adaptive threshold, to reduce the impact of noise onimage segmentation, which is called image preprocessing.In this paper, an remote sensing image segmentation algorithm basedon Markov random field of wavelet domain is proposed,aiming at thecommon multi-resolution segmentation algorithm in remote sensing imageprocessing for each scale of the segmentation process information does notfully take advantage of the image message.The algorithm using wavelet thenon-stationarity of image and multi-scale features,make the Markov modelcharacteristic properties and parameters of the prior model has a variablescale behavior,to better characterize the texture features of theimage.Simulation results show that such a model and algorithms improvethe quality of remote sensing image segmentation.Finally, the paper briefly introduces the evaluation criteria of image segmentation, as well as summarizes the entire thesis work, and putforward in the direction of further research and exploration.