Research On Threshold Segmentation Algorithm Based On OTSU And Maximum Entropy

Image segmentation is a very important technology in the field of image processing,and the segmentation of image directly influences the subsequent results of image processing.Image segmentation has played an important role in monitoring security,medical equipment,industrial production and People's Daily life.In recent years,more and more scholars have been studying new algorithms and applying them to new scenarios.This paper mainly researches and improves the 2-d otsu and maximum entropy algorithm.Large amount of calculation of two-dimensional otsu algorithm segmentation of real-time performance and weak resistance to the noise problem,this paper proposes a new method of improvement,in the heart of the twodimensional histogram diagonal near the main point of division limit in parallel to the diagonal of two parallel lines,used in the area of the two parallel lines with vertical diagonal advanced segmentation,segmentation and fast calculation formula is deduced,and improves the real-time segmentation,and the noise of the two parallel lines outside the error segmentation revised improved integral resistance to build.Finally,the correctness of the improved algorithm is verified by experiment.By maximum entropy image segmentation based on the pixel distribution difference between targets and the background,the effect is clear,and on the basis of the one-dimensional maximum entropy increase to 2 d,of the real time for two-dimensional maximum entropy,the 2-d maximum entropy method through recursive optimization algorithm,improve the real-time segmentation.At the same time,for a better image segmentation,image to improve the noise resistance was divided by two dimensional histogram,because there are a large number of logarithmic in the process of calculation of maximum entropy,and the presence of large amounts of repetitive calculation in the process of calculation,through rapid optimization ability of particle swarm optimization(pso)algorithm,the particle swarm algorithm to join the variation factor in order to avoid in the process of optimization into a local optimum,with maximum entropy as the objective function of the image,search the optimal value in the multidimensional space,each iteration select maximum entropy for latest status,larger particles constantly update location search to the optimal value.