New object level strategy for image fusion quality assessment of high resolution satellite imagery

Nowadays, topographic earth observation satellites provide panchromatic images at a higher spatial resolution and also multi-spectral images at a lower spatial resolution which is rich in spectral information. Image fusion techniques produce new images which inherent the merits of initial panchromatic and multi spectral images. Considering the importance of fusion accuracy on the quality of the next applications, it seems necessary to evaluate the quality of these processed pan-sharpened images. Lots of quality evaluation metrics have been proposed for quality assessment of fused images. These methods are mainly developed on the basis of applying quality metrics in pixel level and to evaluate the fnal quality by averaging computation. In this paper, an object level strategy for fusion quality assessment is proposed. Based on the proposed strategy, image fusion quality metrics are applied on image objects and quality assessments are conducted to inspect fusion quality in those image objects. Results obtained from applying several pan-sharpen QuickBird imagery, clearly show the in-consistency of fusion behaviour in different image objects and the weakness of traditional pixel level strategies in handling these heterogeneities.