Research Of Infrared And Visible Images Fusion Based On CT-domain Hidden Markov Tree Model

With the development of smart substation, condition monitoring of electricity transmission equipment needs to be improved. Because of different imaging mechanism and waveband, the infrared image reflects radiation information of the objective scene but image clarity is lower. The visible image reflects reflection information of the objective scene, and can give a better description of surroundings information. Therefore, fusion of infrared and visible images can make full use of information contained in source images, and can acquire better understanding of the whole scene. Image fusion of infrared and visible image can greatly improve the accuracy of thermal fault location. Several fusion algorithms are studies in this paper, and existing fusion rules are tested and improved. To verify the efficiency, images used in this paper are taken in a500KV substation.To overcome the drawbacks of wavelet transform in dealing with higher dimension signals, fusion algorithm based on CT transform is researched. Fusion rule for the high frequency is Modulus Maxima, while weighted average is for the low frequency. The simulation results show that the fusion result based on CT transform is better than the result based on wavelet transform, reflecting the CT transform of image information ability of characterization.Because of down sampling, CT transform can easily result in frequency alias. Therefore, image fusion algorithm based on NSCT transform is used. In addition, fusion rule based on single pixel is improved, and rule based on the neighborhoods is proposed. The high frequency is based on window variance, while the low frequency is based on window energy. Experiment result proves fusion effect is improved.The fusion algorithm based on CT domain HMT model is focused on in this paper and fusion rule based on combination of CT domain HMT model and edge detection is put forward. In view of the high frequency coefficient contains abundant edge information, fusion rules based on edge detection is researched. Canny criteria is used in the decomposed high frequency coefficients of source image, finding edge feature points and keeping these points as possible fusion coefficient detection in selection rule.Experimental results show that the proposed algorithm provides more satisfying fusion results, in terms of visual effect and objective evaluations, such as entropy, average gradient, mean and standard variance.