Objective Evaluation and Suppressing Effects of Noise in Dynamic Image Fusion

 

Rade Pavlović

Vladimir Petrović

 

In this paper the results of an investigation into the effects of noise on dynamic (video) image fusion performance are presented. We start by creating an extensive multisensor dataset of noise-corrupted videos. Then, we define an objective metric for the evaluation of noisy dynamic fusion N-DQ^AB/F and demonstrate its consistency with visual assessment. The metric to evaluate a wide range of conventional and robust dynamic fusion techniques and strategies for suppressing noise in the fused video on the created dataset is applied. We identify the characteristics of multiresolution pyramid representations and feature selection strategies capable of mitigating the effects of noise on dynamic fusion performance. The paper also shows some relatively simple noise suppression techniques integrated into the fusion process which can yield performance improvements in specially challenging low SNR conditions with very little computational complexity

Key words: image processing, image quality, image fusion, noie measurement, noise filtering, noise suppression.