Article ID Journal Published Year Pages File Type
528813 Information Fusion 2013 12 Pages PDF
Abstract

Given multiple source images of the same scene, image fusion integrates the inherent complementary information into one single image, and thus provides a more complete and accurate description. However, when the source images are of low-resolution, the resultant fused image can still be of low-quality, hindering further image analysis. To improve the resolution, a separate image super-resolution step can be performed. In this paper, we propose a novel framework for simultaneous image fusion and super-resolution. It is based on the use of sparse representations, and consists of three steps. First, the low-resolution source images are interpolated and decomposed into high- and low-frequency components. Sparse coefficients from these components are then computed and fused by using image fusion rules. Finally, the fused sparse coefficients are used to reconstruct a high-resolution fused image. Experiments on various types of source images (including magnetic resonance images, X-ray computed tomography images, visible images, infrared images, and remote sensing images) demonstrate the superiority of the proposed method both quantitatively and qualitatively.

► The similarity between fusion and super-resolution is considered sufficiently. ► A framework of simultaneous image fusion and super-resolution is proposed. ► Sparse representation over learned dictionary is used to implement this framework. ► High resolution fused image is obtained from low resolution source images.

Related Topics
Physical Sciences and Engineering Computer Science Computer Vision and Pattern Recognition
Authors
, , ,