Research Of Image Segmentation Algorithm Based On Ncut

As an important research topic in image processing,image segmentation determines the result and quality of the final image analysis and image understanding.The purpose of this processing method is to divide the target object to be processed into all parts of the uniform property,which is the basic step of the subsequent expansion related calculation and analysis,which is the premise of further studying the area extraction and pattern recognition.In recent years,many scholars have proposed a series of novel and efficient segmentation algorithms.Among these algorithms,normalized segmentation,as a graph-based segmentation algorithm,has been widely used by virtue of its global optimization.Under normal circumstances,the algorithm uses pixels as the processing object to achieve cutting,but the efficiency is lower,the real-time performance is poor,and the noise and other factors are more sensitive.Therefore,it is necessary to improve the traditional Ncut algorithm.The watershed algorithm is based on topological theory and mathematical morphology.The second watershed can effectively reduce the number of subintervals of the target object.Mean-shift algorithms,which are often used to segment images by virtue of their good characteristics,have the drawback of being prone to "premature convergence." The combination of particle swarm optimization can realize adaptive bandwidth selection of the mean shift algorithm.Here,we propose three new improved algorithms.Firstly,we use the above three clustering algorithms to pre-divide the image,and then use the traditional Ncut algorithm to cluster the images.The new algorithm based on this idea can effectively reduce the interference of other factors such as computation and noise,and improve the characteristics of the original algorithm.Finally,we also verify the performance of the improved algorithm experimentally,and compare the running efficiency and segmentation results of the three improved algorithms when dealing with the same image.Through the experimental results,we can see that the three new algorithms have shorter running time and better segmentation than the original method.The Ncut algorithm based on quadratic watershed can effectively preserve the edge information.The Ncut algorithm based on mean shift preserves the color information,and the segmentation result and operation time are satisfactory.The Ncut algorithm based on adaptive mean shift is between complexity and efficiency The good compromise,without morphological processing,to further save time,the algorithm is also the best in real time.