Research On Infrared Image Segmentation Algorithm Based On Active Contour Mode

Infrared imaging technology is widely applied in many areas,such as intelligent driving,security monitoring,artificial intelligence and so on,owing to its advantages of high concealment,strong anti-interference,work day-and-night.Infrared image segmentation which aims to extract the region of interest from the image plays an essential role in infrared target detection and tracking.However,infrared images generally have more complex properties,e.g.,intensity inhomogeneity,blurred boundary,low contrast and high background noise.When the existing segmentation methods directly applied to infrared images,it still cannot obtain ideal results.Active contour model has gained popularity because of its excellent ability to obtain target contours with sub-pixel accuracy.Therefore,based on the selective local or global segmentation(SLGS)models and the characteristics of real infrared images,we design two active contour models which are more suitable for infrared image segmentation.The specific works of this paper are as follows:By comparing the differences between target and background in infrared gray image,entropy image and standard deviation image,an active contour model driven by multi-features information is designed.Firstly,the feature fitting image is constructed by the global feature information and local feature information inside and outside the contour.Then,we calculate the similarity difference between original feature images and feature fitting images to obtain the signed pressure force functions driven by intensity,entropy and standard deviation respectively.Next,afore-mentioned signed pressure functions are added to form a new level set evolution function.Finally,comparative experiments are carried out on real infrared images and the results demonstrate that the presented model outperforms traditional models in terms of precision rate.In order to improve the segmentation efficiency and accuracy of infrared images,an active contour model combined global information with local information is constructed.Firstly,the global term is calculated by the global gray information and the local term is calculated by local entropy,local standard deviation and gradient.Secondly,we design weight coefficient computed by local range to incorporate the aforementioned two terms.Then,the complete signed pressure force function is substituted into the level set function for further evolution.Finally,qualitative and quantitative comparative experiments show that our model has higher accuracy and efficiency in segmenting infrared images.Meanwhile,the performance of image segmentation is almost not influenced by the initial state of level set function.