The variational method was popular and effective for image fusion in recent years, and it could constraint the geometric and gradient structures in original images according to the variational fusion equation with preserving information terms. The sparse representation was widely used for image decomposition and fused image reconstruction by combining sparse coefficients of multi-source images. Therefore, a new variational image fusion method was proposed by adding an approximated term with sparse representation model. What’s more, the sparse representation fusion was improved with the new coefficient fusion rule, and then was joined within the variational fusion frame. The proposed variational fusion method could approximate multiband images and preserve details. The fusion experiments were performed on GF-2 remote sensing images, and compared with original unimproved methods and some usual fusion methods. The results showed that the new proposed method is better than tested methods in accordance with objective evaluations and subjective visual effects.
In this paper, the hardware friendly adaptive support-weight approach is proposed to simplify the weight calculation process of the standard approach, which employs the support region to simplify the calculation of the similarity and uses the fixed distance dependent weight to present the proximity. In addition, the complete stereo matching algorithm and the hardware structure for FPGA implementation compatible with the approach is proposed. The experimental results show that the algorithm produces the disparity map accurately in different illumination conditions and different scenes, and its processing average bad pixel rate is only 6.65% for the standard test images of the Middlebury database, which is approximate to the performance of the standard adaptive support-weight approach. The proposed hardware structure provides a basis for design and implementation of real-time accurate stereo matching FPGA system.