Infrared Target Segmentation Method with Multiscale MRF Based on Data Distribution De-Duced from KLFCV

WEI Kun,ZHAO Yong-qiang,CHENG Yong-mei,GAO Shi-bo,ZHANG Hong-cai
DOI: https://doi.org/10.3321/j.issn:1005-0086.2009.05.023
2009-01-01
Abstract:A new infrared target segmentation method based on infrared multiband images is proposed.First,8-14 μm waveband is subdivided into four bands by light filter,and the single band image is substituted by multi-band′s order to increase the image information.Then,based on the KLFCV fuzzy chuster model,a new data distribution used as likelihood distribution of multi-scale MRF is presented.Finally,the corresponding parameter estimation is given by using EM algorithm.From the result of the experiments on artifact data and real scene image segmentation and comparison with the traditional method,the proposed method displays certain improvement in segmentation integrity and noise reduction.
What problem does this paper attempt to address?