FCM Anti-noise Image Segmentation Method Based On Quadratic Polynomial

Image segmentation is a key link in computer vision and image understanding.The segmentation algorithm should pay special attention to the influence of noise when performing image processing.Segmentation model is the key to improve image segmentation quality.A good segmentation model can make the similarity of pixels classified into the same class as high as possible,and get more accurate segmentation results.When the segmented data points are taken from different curved patches,the existing FCM segmentation algorithm defines the constant as the segmentation center,which often fails to obtain the desired effect,and the noise processing is not accurate enoughThis paper proposes a new FCM algorithm based on quadratic polynomials,which can better distinguish the weak edge regions in the image and has a certain noise immunity.The addition of local sub-segmentation can make the algorithm work more effectively.First,the improved algorithm proposes using a quadratic polynomial surface to define the segmentation center,and dividing the set to which the data point belongs by the algebraic distance from the data point to the segmentation center.The newly defined segmented surface is more inclusive of the features of the pixels,and can effectively deal with the problem that the gray level of the image is much larger than the number of cluster centers.Second,based on the quadratic polynomial surface as the center of segmentation,a new fuzzy factor is designed.In the calculation process,the improved algorithm uses the deviation value to represent the difference between the average algebraic distance of the neighborhood points and the algebraic distance of the central pixel.The influence of the neighborhood point on the center point can be measured by calculating the deviation value.In the presence of noise,the offset value can offset the influence of noise in the calculation process,so the improved algorithm has an effective improvement in noise resistance.Third,select a local window on the edge of the global segmentation result to perform local window segmentation,and use depth-first search to select the window to perform local segmentation processing on the image again.The gray level of the pixels in the local small window is much smaller than the gray level of the pixels in the entire image.This is equivalent to using a cluster center that is more in line with local information to segment the local small window and optimize some misclassified pixels.Get the final segmentation result.The experimental results show that in the final segmentation result of medical images with 5%noise,the segmentation accuracy can reach more than 96%,and the partition coefficient Vpc is increased by 0.14.Therefore,the algorithm can obtain a membership matrix with less ambiguity and obtain more reliable segmentation results,and can effectively eliminate the effects of noise and retain image details.This algorithm is effective not only for medical image segmentation,but also for natural and synthetic images,which provides a good basis for subsequent algorithms for image processing.