Sparse based simultaneous fusion and super resolution of multi-modal images

Image fusion and Super Resolution techniques are often desirable in many imaging applications for further processing, classification and analysis of images. Higher resolution fused images often provides more detailed information than its low resolution counterpart. In this paper, we present a novel sparse based simultaneous Fusion and Super Resolution of multi-modal images. The source image pairs are initially separated into low and high frequency images. Sparse vectors for these images are obtained using learned dictionaries. The sparse coefficients are fused and with fused coefficients the enhanced higher resolution image is retrieved back. The proposed method is validated using five popular metrics and a comparative analysis of the method with various dictionary sizes is also carried out.