Abstract: For forestry robots, it is easily to be affected by slightly irregular movement such as sliding, collision of ground obstacles and etc., which will lead to motion blurred images acquired by robot camera, and will greatly influence the results of subsequent information extraction. In order to solve this problem, the forestry blurred image restoration method based on synthetic regularization is proposed. Firstly, a cost function containing [L1/L2] norm regularization is set up to solve motion blur kernel. Then, another cost function consisting of image gradient prior regularization and sparse regularization are built to acquire the target clear image. The introduction of [L1/L2] norm regularization and image gradient prior regularization can make up blocking problems caused by sparse representation regularization, thus satisfactory results are obtained. Finally, the experimentd results of the synthetic and the natural real motion blurred images verify the effectiveness of the proposed method.

摘要： 林业机器人在林业环境中进行作业时，很容易因为滑动、地面障碍物的碰撞等原因发生小幅的无规律运动导致机器人相机采集的图像发生运动模糊，对后续图像信息提取造成很大的影响。针对这一问题，提出了林业运动模糊图像复原的融合正则化方法。先建立包含[L1/L2]范数正则项的代价函数，求解运动模糊核。再通过图像梯度先验正则项及稀疏正则项构建代价函数，对清晰图像求解。引入的[L1/L2]范数正则项及图像梯度先验正则项对稀疏表示正则项容易产生块效应的问题进行了弥补，因而获得了令人满意的效果。对人工合成的运动模糊图像和自然条件下真实运动模糊图像进行的实验验证了该算法的有效性。

关键词: 运动模糊图像, 图像盲复原, 稀疏表达, 梯度约束, 归一化范数