The Optimization Of Otsu Algorithm

Image segmentation is the basic process of modern digital image processing technology.It is used to pretreat the advanced image processing technology of artificial intelligence.Over the years,countless methods have been proposed to improve their accuracy and time.Otsu is a kind of the classical algorithm which realizes image segmentation by selecting threshold.Because of its convenient and practical features,it has been developed rapidly and dramatically.The dimension of algorithm has been improved from one dimension to multi-dimension,and the threshold value of algorithm has been extended from one to more.However,the time required to complete the segmentation is geometrically increasing,and real-time segmentation is the main constraint in the practical application of algorithm.Therefore,to the Otsu algorithm,in the premise of ensuring the segmentation accuracy of the algorithm how to reduce the algorithm's computation time and improve the performance of the algorithm is the key point.On the basis of carefully studying the2 d Otsu algorithm(2-Otsu)and three-dimensional Otsu algorithm(3-Otsu),the specific work is as follows:1.After learning the standard FA and thinking about its own strengths and weaknesses,the search rate of the standard FA is higher,but there are problems that it is prone to local optimal and poor convergence performance.In order to improve the FA performance of the standard,a step length adjustment function was introduced in this paper,and the step length matched with the number of iterations and convergence was obtained adaptively.2.An image segmentation algorithm for improving FA optimization 2-Otsu is proposed.Matlab simulation results show that the proposed algorithm reduces the computational complexity of the algorithm on the basis of ensuring the segmentation effect is roughly constant;and under the premise that Shannon entropy and regional contrast are basically consistent,the proposed algorithm reduces the time required to complete the algorithm,which is a better comprehensive performance algorithm.3.By the way of the projection,the 3-Otsu is decomposed into three 2-Otsu,and the segmentation results are combined and optimized as outputs.Matlab simulation results show that in the time,the algorithm is nearly 30 times less than 3-Otsu,and slightly more than 2-Otsu but its value is still small;in the anti-noise performance,especially with mixed noise,the algorithm is superior to the above two algorithms.