Based On Statistic Image Segmentation Method Research And Application

Image segmentation is a very important research topic in image processing,pattern recognition and artificial intelligence areas, image segmentation based on thestatistics combined mathematical model in image segmentation theory locationalgorithm. Remote sensing image is usually manifested as: a large amount ofinformation, the fuzzy boundary, a lot of gray level, complex target structure andcharacteristics, because these properties making the model is not completely reliableguide remote sensing image segmentation, to some extent,which hindered thesegmentation technology in the remote sensing application field. Due to the largeamount of data, high resolution, no injury to human body features, Nuclear magneticresonance images was widely used in the medical field. Because of very high valuemedical image segmentation in recent years, brain MRI image segmentation hasbecome a research focus in the medical image segmentation application. This paperfocuses on the image segmentation method based on statistics, and these methods areapplied to the segmentation of remote sensing image and magnetic resonance imagesegmentation, through experimental comparison of remote sensing images andmagnetic resonance images find for a superior algorithm on the basis of these twotypes of images.The main contents of this paper include the following:1) Statistical image segmentation method research. The existing statistical imagesegmentation is discussed, including OTSU method, moment invariants algorithmand maximum entropy segmentation algorithm, they provided reference for furtherstudy in the neighborhood.2) The characteristics of MRF model in image segmentation algorithm. First ofall, we introduced the two-dimensional MRF model and its related concepts, thenillustrated the distribution of Gibbs and MRF equivalence relations, proposed MRF-MAP framework, it was commonly used methods on the basis of MRFapplication, finally introduced the application of MRF image segmentationalgorithms, including the ICM algorithm and SA algorithm. In view of the features oftraditional MRF potential energy function is too simple, according to the context andthe Euclidean distance to improve the potential energy function, through the result ofexperiments show their effectiveness.3) Introduction the features of remote sensing image and magnetic resonanceimage. Statistical image segmentation algorithms are applied to remote sensing imageand magnetic resonance images, these statistical image segmentation algorithmsinclude K-means algorithm, the traditional MRF algorithm and improved potentialenergy function of MRF algorithm. Through experimental analysis contrast we candraw the improved potential energy function of the MRF algorithm for magneticresonance images and remote sensing image segmentation effect best.