Color correction for multiview images is a fundamental yet challenging task to generate a color-consistent composite image and has important applications in remote sensing and 3-D vision. In this article, we propose an effective fusion algorithm for multiview color correction using a graph-based adaptive minimum weight-first-then-larger outdegree (AMWLO) strategy to select the next target image. Initially, the given multiview images are stitched using an existing image stitching method. First, we take a modified color distance (MCD) metric to estimate the residual of the overlapped region of each image pair as the weight of the edge. After constructing the graph model of the given multiview images, an AMWLO approach is proposed to select the first target image to be performed for color correction. Then, integrating the histogram equalization (HE) and the joint bilateral interpolation (JBI), the proposed fusion-based method is used to correct the color for the selected target image. After that, the weights between the color-corrected target image and its overlapped images are updated. The above AMWLO-based target image selection, color correction for the selected target image, and edge weight update are repeated until the color correction for all multiview images is completed. To better evaluate the quality performance of the color-corrected stitched multiview images, some modified quality assessments are proposed by taking the overlapping area factor into account. Based on large-scale testing multiview image datasets, the comprehensive experimental results demonstrated the quantitative and qualitative quality merits of our algorithm relative to the state-of-the-art methods.

中文翻译：

---

通过最小权重先大出度目标图像选择进行基于融合的多视图色彩校正


多视图图像的颜色校正是生成颜色一致的合成图像的一项基本但具有挑战性的任务，并且在遥感和 3D 视觉中具有重要应用。在本文中，我们提出了一种有效的多视图色彩校正融合算法，使用基于图的自适应最小权重先大出度（AMWLO）策略来选择下一个目标图像。最初，使用现有的图像拼接方法来拼接给定的多视图图像。首先，我们采用修正的颜色距离（MCD）度量来估计每个图像对的重叠区域的残差作为边缘的权重。在构建给定多视图图像的图形模型后，提出了 AMWLO 方法来选择要执行颜色校正的第一个目标图像。然后，集成直方图均衡（HE）和联合双边插值（JBI），所提出的基于融合的方法用于校正所选目标图像的颜色。之后，更新颜色校正后的目标图像与其重叠图像之间的权重。重复上述基于AMWLO的目标图像选择、对所选目标图像的色彩校正以及边缘权重更新，直到完成所有多视图图像的色彩校正。为了更好地评估色彩校正拼接多视图图像的质量性能，通过考虑重叠面积因素，提出了一些改进的质量评估。基于大规模测试多视图图像数据集，综合实验结果证明了我们的算法相对于最先进的方法的定量和定性质量优点。