Abstract
The problem of image segmentation can be reduced to the clustering of pixels in the intensity space. The traditional fuzzy c-means algorithm only uses pixel membership information and does not make full use of spatial information around the pixel, so it is not ideal for noise reduction. Therefore, this paper proposes a clustering algorithm based on spatial information to improve the anti-noise and accuracy of image segmentation. Firstly, the image is roughly clustered using the improved Lévy grey wolf optimization algorithm (LGWO) to obtain the initial clustering center. Secondly, the neighborhood and non-neighborhood information around the pixel is added into the target function as spatial information, the weight between the pixel information and non-neighborhood spatial information is adjusted by information entropy, and the traditional Euclidean distance is replaced by the improved distance measure. Finally, the objective function is optimized by the gradient descent method to segment the image correctly.
Highlights
In recent years, clustering technology has played an important role in remote sensing image segmentation
(1 + α) um ij where c is the number of clusters, n is the number of pixels in the image, uij denotes the membership c uij = 1, m degree of x j in the ith cluster, has a value inside [0,1] and t satisfies the condition 0 ≤ uij ≤ 1, i=1 is the fuzzy weight index and is generally a value of 2, x j − vi represents the Euclidean distance from the jth pixel x j to the ith clustering center vi, NR is the window cardinality, xr denotes the neighborhood pixel set centered on the jth pixel x j, α is the influence factor of neighborhood spatial information on the center pixel
Where SA represents the proportion of pixels in the region detected by the segmentation algorithm in the whole region and CS is a measure of similarity
Summary
In recent years, clustering technology has played an important role in remote sensing image segmentation. When the image noise is relatively serious, the neighborhood information of the pixel may be polluted, so the neighborhood information based on the local space of the image cannot play an active guiding role in the image segmentation, making the fuzzy clustering algorithm that integrates the local space information unable to meet the requirements of high-precision image segmentation. To solve this problem, Zhao et al [16] proposed a fuzzy c-means clustering algorithm based on non-local spatial information (the FCM_NLS algorithm). The segmentation results of different images show that this algorithm can achieve better segmentation performance under intense noise
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
