Research On Gray Level Uneven Image Segmentation Method Based On Level Set

Image segmentation is based on texture features,gray information,color and shape,and is divided into regions with corresponding features,and then further image processing from the region of interest is an important step in the image processing process.Among all the segmentation algorithms,the level set algorithm relies on the idea of dynamic evolution of contour curves,and is widely used due to its superior topology and powerful mathematical foundation.However,due to the influence of different complex scenes,the image exhibits different degrees of gray unevenness,which brings great challenges to segmentation.At present,the accurate segmentation of grayscale inhomogeneous images based on the level set method is currently a research hotspot.This dissertation mainly studies the segmentation based on the level set method on the gray scale inhomogeneous image.The defects and shortcomings of the local binary fitting model segmentation are improved from two angles.The main work is as follows:(1)For the local binary fitting model,the slow evolution of the curve is slow and the weak edge segmentation is not ideal.A local binary fitting energy model based on local introduction of global information is proposed.The method accelerates the evolution of the contour by introducing a local global force in the uniform region of the gray scale,and uses the gray uniform judgment function to determine the uniformity of the gray distribution of the region around each point on the curve,thereby determining the use range of the local global force,and calculating It moves in the direction of energy minimization,achieving fast and accurate segmentation of the target area.(2)For the local binary fitting model,the manual initialization is required to be sensitive to the initial contour position and size.The improved local binary fitting model that introduces local global information does not overcome the problem of sensitivity to the initial contour.A local binary fitting model introducing fuzzy ideas is proposed.The method preprocesses the image by fuzzy C-means clustering method,selects the target of interest,converts its target membership value into the initial value of the level set function that the model adapts,and then introduces local binary fitting by introducing local global information.The model evolves from the initial value to the completion of the segmentation to achieve the final segmentation goal.Finally,through simulation experiments,the comparison between the correct measurement of segmentation,error measurement and time used for segmentation shows that the improved algorithm proposed in this dissertation has better performance than the local binary fitting model.Figure[20] table[4] reference[65]...