Objective Evaluating Method to Fusion Image Quality Based on ANN

Research on quality evaluating of fusion images is meaningful to improve the registration technology and fusion algorithms. The objective assessment metrics based on fusion image itself can not show comprehensively the fusion image quality, which could be incongruent against human eye response. Provide a new assessment method for fusion image quality by means of building artificial neural network model. The metrics including mean value, standard deviation, gradient is choosed as the input neural cell. A hidden layer is designed to carry out assorting performance. And the artificial neural network obtain the image quality assessing mapping functions and to classify the training samples into different types by means of the supervised learning. The experiment shows the noteworthy concordance between the simulation result and human eye response to identifying samples. Compared with single evaluation indexes, the new quality evaluation model can show effectively the subjective response of human eye to fusion image.