HIDE 图像去模糊数据集

HIDE 数据集主要针对动态去模糊问题提出。由成像设备和场景之间的相对运动引起的多种模糊，主要包括相机晃动和物体移动。为了充分捕捉由被动设备干扰和主动行为引起的动态模糊，HIDE 数据集应运而生，用以覆盖更大范围和更近距离的场景，解决人体可察觉的运动去模糊问题。

The HIDE dataset is primarily proposed for dynamic deblurring tasks. Multiple types of blurs, mainly camera shake and object motion, arise from relative movement between imaging devices and their corresponding scenes. To fully capture dynamic blurs caused by both passive device disturbances and active human behaviors, the HIDE dataset was developed to cover a broader range of closer-range scenes and address human-perceivable motion deblurring problems.

搜集汇总

背景与挑战

背景概述

HIDE数据集是一个针对动态去模糊问题的图像数据集，包含8,422个模糊-清晰图像对和65,784个密集标注的前景人体边界框，主要用于解决由相机晃动和物体移动引起的动态模糊问题。数据集大小为15.81 GB，覆盖了更广泛和更近距离的场景，适用于图像处理和图像理解等领域的研究。