Research On Image Segmentation Quality Evaluation Based On Combination Weighting

Image segmentation is a key step in the process of image processing to image analysis.The quality of image segmentation directly affects the follow-up links,like target recognition and so on.Therefore,the research on image segmentation has been one of the hot topics in image engineering.At present,many segmentation algorithms have been proposed,but it is still a difficult problem to evaluate whether the algorithm is superior to other algorithms.To overcome this difficulty,it is necessary to study the method of image segmentation evaluation.The evaluations about segmentation algorithms include the evaluation of segmentation quality,noise immunity and computational complexity.In this paper,a comprehensive evaluation model is established as for image segmentation quality evaluation.The main contents of this paper include the following three aspects:1.Aiming at the problem that the current image segmentation evaluation index can not reflect the segmentation results well,a gray evaluation model of combination weighting is proposed.First of all,we choose three criteria: Probabilistic Rand Index,Global Consistency Error and Variation of Information,to evaluate the quality of image segmentation.Secondly,we propose a subjective and objective combination weighting method combining Delphi,Forced Decision and entropy method.Finally,the proposed model is used to make a comprehensive evaluation of the test images.The experimental results show that the proposed evaluation model is more in line with the subjective evaluation results and the real ground results.2.Based on the maximum entropy multi-threshold segmentation algorithm,a multi-threshold image segmentation algorithm based on Flower Pollination Algorithm is proposed.It is verified that the Flower Pollination Algorithm is better than Genetic Algorithm and Shuffled Frog Leaping Algorithm in multi-threshold segmentation.3.The gray evaluation model of image segmentation based on combination weighting established above is used to compare the segments,which are processed by the maximum entropy multi-threshold algorithm based on Flower Pollination Algorithm,Genetic Algorithm and Frog Leaping Algorithm respectively.Experiments show that ranking results by the model are consistent with the ranking results of maximum entropy,which further validates the validity of the model and the superiority of multi-threshold image segmentation algorithm based on Flower Pollination Algorithm.