Image Fusion Quality Assessment Based On Structural Similarity

Assessment of image fusion quality is the important part of image fusion, and itis also one of the most difficult problems in this field. Assessment of image fusionquality affects the choice of algorithms and parameters. The metric for assessment ofimage fusion must be simple, reasonable and according with Human VisualSystem(HVS). Structural similarity points out a new way for the research of imagefusion quality. In this thesis, we studied quality metrics for image fusion based onstructural similarity, the main work includes:1. Studied the structural similarity, which was based on the image structuralinformation. Proved the availability of structural similarity for assessment of imagefusion quality by experiments and theory analysis. Proposed several quality metricsfor image fusion, which were based on average structural similarity, and these metricshave better effect.2. Human have the characteristic of being sensitive to edge. Combining the edgeinformation of image, the quality metric for image fusion based on gradient structuralsimilarity was proposed in this thesis. In experiments, it was proved that thisalgorithm is more according with Human Visual System than algorithms in existence.3. Analysis the characteristics of Human Visual System. Chose the suitable modelsof brightness perceptive and contrast sensitivity function (CSF). Combining the HVSmodel, quality metrics for image fusion based on structural similarity was proposed,and the availability of this algorithm has been proved by experiments and theoryanalysis.