Reasearch On Superpixel Segmentation Algorithm Based On Boundary Preservation

Recently,with the popularization of mobile terminal devices,images have gradually become the most important information carriers in daily life.The development of optical imaging equipment has greatly improved the image quality,and the number of pixels in the image has also been increasing.Generally speaking,the complexity of related algorithms in the field of digital image processing is directly proportional to the number of pixels in the image,so the increase in the number of image pixels will seriously affect the execution efficiency of the image processing algorithm.In natural images,there is usually a large amount of redundant information,and many adjacent pixels have similar characteristics such as color and texture.The superpixel generation algorithm uses a superpixel to represent these pixels with similar characteristics.Using several superpixels to replace all pixels in the image as the basic unit of image processing can significantly improve the efficiency of image processing algorithms.Therefore,superpixel segmentation is widely used in many fields such as image segmentation and target detection.In these application fields,the quality of superpixels will determine the final results of subsequent algorithms.Generally speaking,higher-quality superpixels have regular shapes and can closely fit the boundaries of objects in the image.However,it is not suitable for existing superpixel generation algorithms.It is difficult to meet this requirement.This is because when the superpixel is close to the boundary of the object,its shape will become more irregular.On the contrary,to generate a regular superpixel means to ignore some of the details in the image,so It will cause the degree of fit of the superpixel boundary to decrease.How to reasonably use the various features of the pixels to make the generated superpixels close to the image boundary while maintaining regularity has become a research difficulty in the field of superpixel segmentation.In response to the above problems,this paper proposes a superpixel segmentation algorithm based on boundary preservation.First,the algorithm uses side window filtering to process the image.This step smoothes the texture area appropriately to ensure the superpixel shape generated in the texture area.More regular.In addition,this paper designs a new distance measurement function,which can assign different weights to its color distance items and spatial distance items according to different pixels during distance measurement,so that the superpixels fit the boundaries in the image boundary area and are flat.The shape of the area is regular.The distance measurement function also takes into account the pixel information of the linear path from the pixel to the cluster center,and avoids the category error division caused by only the local information of the pixel for clustering.Finally,this paper designs a cluster center update strategy,using only the weighted average of some reliable pixels in the superpixels as the new cluster center,which reduces the interference of abnormal pixels with low similarity to the cluster center.To make the update of cluster centers more accurate.This paper verifies the effectiveness of the proposed algorithm through comparative experiments.This paper uses BSD500 as the test image set and compares it with the current mainstream superpixel segmentation algorithms.The results show that the algorithm proposed in this paper has certain advantages both in terms of vision and related quantitative indicators.