Research On The Algorithm Of Neuron Structure Extraction In Microscopic Images

Neuron is the basic unit that constitutes the structure and function of human nervous system.The correct segmentation of neurons is the basis of constructing the three-dimensional structure of neurons,so segmenting neuron images correctly has an important practical significance.Two photon laser microscopy is one of the main devices to obtain the microscopic images of neurons,However,image degradation and blur phenomenon may exist during the procedure of confocal microscopic imaging,and due to the uneven distribution of neuron image’s gray pixels,the low contrast of target and background,and the noise pollution,correct segmentation in neuron image becomes extremely difficult.In order to solve this problem,based on relevant work from many literatures and the features of neurons,this paper puts forward two methods for neuron image segmentation.The research content of this paper is as follows:Firstly,the pre-processing of neuron image.By analyzing the common methods of image enhancement,In order to solve the problem that the object and background of the neuron image have low contrast and poor connectivity,this paper mainly introduces the Hession matrix filter and the steerable filter which are suitable for linear image enhancement.Because the direction of traditional steerable filter is rough,it can not accurately reflect the structure of neurons.In this paper,an improved method of steerable filter under the constraint of Canny like criterion is given.The experimental results show that the Hession matrix filter and the improved steerable filter can effectively enhance the neuron image.Secondly,neuron image segmentation.This paper studies two kinds of neuron segmentation methods.Because of the similarity between the neuron image and the retinal image,and the problems of weak connectivity of neuron images,an improved method to extend the tracking range is proposed based on the research of retinal tracking algorithm,the experimental results show that the proposed method is effective for neuron image segmentation.In order to further enhance the segmentation effect,considering the advantages and disadvantages of traditional threshold segmentation methods in neuron image segmentation,a new neuron image segmentation method is proposed: neuron image segmentation algorithm based on connected domain.In this new algorithm,participate in the threshold of the object is no longer a single pixel,but the connected domain.Finally,evaluation of the experimental results.At present,there are few researches on the evaluating methods of the neuron image segmentation algorithms.For most cases,theexperimental results are analyzed and evaluated by observing with raw eyes.In order to objectively evaluate the result of the experiment,in the absence of contrast template library,this paper presents a neuron image segmentation evaluating method based on pixel.This method can evaluate the result of the experiment without contrast template library.The method of neuron image evaluation based on pixel is used to evaluate the performance of the neuron image segmentation algorithm.The experimental results show that both algorithms proposed are effective for neuron image segmentation.And that the neuron image segmentation algorithm based on connected domain is more effective.