Abstract: For the compressive sensing coding of an aerial image set, the current methods only allocate a regular sampling subrate for each image in the image set. Without considering the different characteristics of image scenes and the overall quality of an image set, the regular allocation mechanism is difficult to properly utilize the limited sampling resources. Under the constraint of total sampling resources, it is still an open issue how to efficiently provide a certain sampling subrate for a different image in an aerial image set. To assess the overall quality of an image set, this paper firstly proposes a hybrid quality metric according to the general requirements of an aerial image set. By evaluating the relative complexity of different aerial images, this paper establishes an image-level variation model, and then proposes a model-guided subrate allocation algorithm for different images in an image set. The experimental results show that compared with the existing methods, the proposed algorithm can significantly improve the overall quality of each aerial image set.

Key words: compressive sensing coding, aerial image set, hybrid quality, sampling subrate

摘要： 对于一个航拍图像集的压缩感知编码，现有方案只能采用固定的测量分配机制对其中每幅图像进行压缩采样，未考虑图像之间的差异性以及图像集的整体重构质量，因此难以充分利用有限的采样资源。在总的采样资源约束下，如何为航拍图像集中不同图像分配合理的采样率是一个需要解决的问题。首先，根据航拍图像集的通用需求提出了图像集复合质量的评价指标，用以计算图像集的整体重构质量；随后，根据图像集中不同图像的相对复杂度建立了一个图像方差模型，并基于该模型提出了一种图像集的压缩采样分配算法。实验结果表明相比于现有方案，在相同的采样资源约束下，所提算法有效地提升了航拍图像集的整体重构质量。

关键词: 压缩感知编码, 航拍图像集, 复合质量, 采样率