目的 色彩纠正和图像融合是生成高质量全景场景图像的关键技术。色彩纠正中参考图像的选择、以及图像融合算法，决定着所生成全景图像的质量和速度。现有方法在确定一幅图像是否适合作为参考图像时，需要遍历所有其它图像，计算其作为参考图像进行色彩纠正的效果，复杂度高，速度慢；在图像融合时存在融合质量与融合速度之间的矛盾。因此，如何快速生成高质量的全景图像就成为全景场景再现的主要诉求。方法 为此本文提出优化的参考图像自动选择的色彩纠正方法和基于重叠区域划分的分区融合方法。针对参考图像选择算法复杂度高的问题，本文根据图像质量与稳定性通常呈反比关系的事实，采用贪婪策略，选择质量最差的图像在色彩纠正前后的相似度，作为是否选择当前图像作为参考图像的评价指标，在保证参考图像满足色彩纠正需求的前提下，大幅降低计算复杂度。针对融合质量与融合速度之间的矛盾，提出分区融合：将重叠区域划分为接缝区域和非接缝区域，利用泊松融合的接缝不可见性和线性融合实现速度快的特性分别对接缝区域和非接缝区域进行融合，既保证融合的质量，又加快融合速度。在此基础上，加入简单点光源，解决上述过程产生的光线一致性问题，进一步提高图像质量。结果 采用主观和客观相结合的方法对结果进行评估。主观方面，本文算法生成的全景图像色彩基本实现平滑过渡且图像原始信息保留完整。客观方面，色彩纠正前后图像的结构相似度（SSIM）控制在了0.850.99之间，时间复杂度由原来的降低到；分区融合后图像信息熵接近于泊松融合，但时间消耗降低72%。采用基于PC端的问卷调查法和OG-IQA算法将本文算法与PTGui、OpenCV、Xiong方法生成的全景图质量进行对比，在大多数情况下本文算法均优于上述算法。结论 实验表明，本文算法适用于多种场景，在保证目视效果良好的前提下，时间消耗降低，可广泛应用于医学、数字旅游、遥感等领域。
Objective Panorama scene reproduction, widely used in medicine, tourism, remote sensing and photography, is the technique for producing a large or wide-angle scene with multiple overlapping images. Color correction and image blending are key issues in the generation of high-quality panoramas. The generation efficiency and quality are primarily determined by selecting the reference image for color correction and the image blending algorithm. To determine a reference image, the state of the art methods need to compare the similarity of all target images, which is computational complex with poor real-time responsiveness. Also, there is a contradiction between quality and speed in image blending. Therefore, how to quickly generate a high-quality panoramic image becomes the primary concern for reproducing panorama scene.Method The key task for the panorama stitching technique is to find the optimal seams in the overlap region of the source image, merge them along the seam, and minimize the seam artifact. This paper presents an efficient method for selecting a reference image for color correction, and a partition blending method which differentiates the overlapping area. Due to the difference in camera equipment, shooting angle and shooting time, the color and illumination of images are inconsistent, which will affect the visual quality of the panorama. Color correction is used for reducing the color difference between images and can accelerate the process for finding the optimal seam and image blending. Color correction uses a reference image to adjust the color style of other images. In other words, the reference image determines the quality of the final panorama. When the quality of the selected reference image is poor, the panoramic image may suffer from blur, inappropriate brightness, and low contrast. According to the fact that the quality of an image is usually inversely proportional to the stability of an image, a greedy strategy is adopted to select the best reference image to reduce the computational complexity. The worst quality image is selected as the baseline, judged by the relative standard deviation of the image pixels of the adjacent images. The similarity between the original baseline and the corrected baseline is used to judge whether an input image is appropriate to act as the reference image. In such a way, the complexity for selecting reference image is significantly reduced, while guaranteeing the need for color correction. An effective color correction method can achieve a smooth transition of the panoramic image. However, there is still a phenomenon of unnatural transition along the seam. Image blending is necessary to further conceal the artifacts, and Poisson blending and linear blending are usually used to do so. Linear blending is simple and fast. However, the weakness is that the stitching artifacts may be visible after the blending. Compared to Linear blending, Poisson blending is more effective while it is more time-consuming than the former. Aiming at these problems, the partition blending is proposed: the overlapping region is divided into the seam region and the non-seam region, and we perform Poisson blending in the seam region and linear correction in the non-seam region to get high-quality image efficiently. A simple point-light source is added to further solve the problem of light inconsistency generated by the above processes and improve the quality of the panorama.Result Both the subjective and objective evaluation show encouraging results for the proposed methods. For the subjective evaluation, the methods can produce a panoramic scene with consistent color styles and keep original details. For the objective evaluation, the structural similarity (SSIM) of the image after color correction is controlled between 0.85 and 0.99, the time complexity is reduced tofrom the original, the image information entropy is close to the Poisson blending after the partition blending, and the time consumption is 72% lower than its original. In addition, we use the PC-based questionnaire method and OG-IQA algorithm to compare the quality of the panorama generated by PTGui, the OpenCV, Xiong’s method and our proposed method. The results show that our proposed method performs the best in most cases.Conclusion The experiments demonstrate that our proposed methods work well on various scenarios. The time consumption is reduced while ensuring good visual effect, and the methods can be widely used in medicine, digital tourism, remote sensing and other fields.