Error Compensation-Based Fusion Algorithm for Drone-image Color Correction

Color correction for drone images, which are captured by unmanned aerial vehicles, is an essential task in drone-image based intelligent applications. Different from existing methods, in this paper, we propose an effective error compensation-based fusion algorithm for drone-image color correction. In our first novelty, an image feature matching method is performed on all image pairs to determine their matched feature point pairs and the overlapping areas. To correct color for each target pixel in the overlapping area, we propose an error compensation-based fusion method. The proposed fusion method combines the joint bilateral interpolation (JBI), which works on the color differences of the matched feature point pairs in the overlapping area, and the histogram equalization (HE), which works on the whole source and target pixels in the overlapping area, such that the overall errors caused by JBI and HE can be minimized. In our second novelty, to better correct color for each target pixel in the non-overlapping area, a boundary reference interval-based fusion method is proposed by using the color differences on the boundary and the color-corrected target sub-region in the overlapping area. Based on seven challenging datasets, comprehensive experiments have been carried out. In terms of thorough quality metrics, the experimental data demonstrate the substantial quantitative and qualitative quality improvements of our algorithm when compared to state-of-the-art methods. The source code of our algorithm is available at https://github.com/ivpml84079/EC-Based-Color-Correction.git.