Research On Improved DNLS Image Segmentation Algorithm

With the development of computer technology,image has become an important method to obtain and transmit information.Image segmentation is the foundation of image understanding and analysis and has been widely adopted in the fields of transportation,meteorology,medicine,military and agriculture.However,image segmentation is still facing a number of difficulties and challenges due to the different characteristics of images,such as uneven illumination and gray scale distribution.In order to resolve these problems,the disjunctive normal level set(DNLS)algorithm is adopted in image segmentation research.The proposed algorithm has lower requirements on the homogeneity of pixels and the step size is not constrained by the number of standard convergence conditions,therefore,it has better time performance and better segmentation effect.However,there are still some aspects to be improved in DNLS algorithm,such as incomplete edge fitting and inaccurate segmentation of complex images due to lack of topological property.Aiming at these problems,this paper combines with other methods to improve the DNLS algorithm,in the meantime,the improved DNLS image segmentation algorithm is combined with threshold segmentation and applied to vegetation coverage area ratio detection.The main contents of this paper are as follows:(1)Combining with the method of Bidimensional Empirical Mode Decomposition(BEMD),a DNLS image segmentation algorithm based on BEMD is proposed.Firstly,the image is decomposed into several intrinsic mode functions(IMF)and a margin by adopting the BEMD principle to make full use of the valid information of pixel value.Secondly,the IMF component extracted by BEMD is given weight and the energy function of DNLS model is rewritten to obtain the improved DNLS algorithm.The segmentation experiments are conducted with other algorithms under the same environment to verify the effectiveness of the algorithm.(2)Combining with Gaussian Markov random fields(GMRF)and HSV(Hue,Saturation,Value)color model,a DNLS color image segmentation algorithm based on GMRF is put forward.Firstly,the color channel information of H,S and V of the image is extracted.Then,the GMRF model is adopted to describe the texture features of different color channels.Finally,the texture information and color information of the image are weighted and the energy function of the DNLS model is rewritten to obtain the improved DNLS algorithm.The segmentation experiments are conducted with other algorithms under the same environment to verify the effectiveness of the algorithm.(3)Combining with the threshold segmentation method,a method of vegetation coverage area ratio detection based on DNLS color image segmentation algorithm of GMRF is put forward.Firstly,by comparing two improved algorithms,DNLS color image segmentation algorithm based on GMRF is adopted for pre-segmentation.Secondly,the OTSU algorithm is used for threshold segmentation.Finally,the final segmentation results are obtained by morphological processing and the area ratio parameters are extracted.The segmentation experiments are conducted with other algorithms under the same environment to verify the effectiveness of the algorithm.