Research on objective evaluation model for fusion image of multi-source detector

Abstract

Aiming at the problem of the lack of comprehensive and objective evaluation methods for near-infrared and low-light-level multi-band fusion images, this paper proposes a multi-dimensional fusion image quality evaluation model that conforms to the interpretation habits of the human visual system. Firstly, the multi-source fusion image quality of low illumination CMOS, ICCD, EBAPS devices and near-infrared InGaAs detector is subjectively evaluated. Secondly, 14 classical evaluation indexes such as information entropy, mean gradient, mutual information and structure similarity are combined. The parameters of each index are dynamically adjusted, and finally an objective evaluation model is established based on the linear regression model. The SROCC, KROCC, PLCC and RMSE values of the optimized objective evaluation model and subjective evaluation score are 0.88, 0.72, 0.89 and 0.38, respectively, indicating that the accuracy and stability evaluation effect of the objective model is better. It solves the limitation of single factor evaluation index and realizes the comprehensive and objective quality evaluation of multi-source detector fusion image, which provides theoretical support and decision basis for the selection of night vision camera in complex illumination environment and the data fusion of multi-detector in low illumination condition. < <