翻译成英文：深度图像是计算机视觉和机器人领域中的重要数据类型可以提供三维场景的距离信息。由于传感器的限制或场景的复杂性原始深度图像常常出现缺失深度信息的问题。本文提出了一种新的信息熵引导的深度修复策略该策略可以有效地修补原始深度图像中缺失的深度信息。首先原始深度图和彩色图像被进行预处理为后续步骤提供连通孔洞图像和灰度图像。然后评估无效点的填充优先级在这个过程中引入了信息熵的概念具体参考了无效点所在

Depth image is an important data type in the fields of computer vision and robotics, which can provide distance information of three-dimensional scenes. Due to sensor limitations or scene complexity, original depth images often suffer from missing depth information. This paper proposes a new entropy-guided depth restoration strategy, which can effectively repair the missing depth information in original depth images. First, the original depth image and color image are preprocessed to provide connected hole maps and grayscale images for subsequent steps. Then, the filling priority of invalid points is evaluated, during which the concept of entropy is introduced, specifically referring to the depth information of the valid neighborhood where the invalid points are located. Next, the prediction of invalid point depth values is accomplished based on the guidance of color and gradient information. In the prediction step, color information ensures overall accuracy, while gradient information is utilized to further improve the precision of restoration. Finally, comparative experiments are conducted on the Middlebury dataset. Compared with other intelligent methods, the proposed method demonstrates better robustness and accuracy, providing new ideas for related research in the field of depth image processing.