光学学报, 2020, 40 (19): 1928001, 网络出版: 2020-09-29
基于离散余弦变换的无人机耀斑图像恢复算法 下载: 934次
Unmanned Aerial Vehicle Glint Image Restoration Algorithm Based on Discrete Cosine Transform
摘要
针对当前遥感图像耀斑恢复算法存在的耀斑增益计算误差大,需要近红外波段信息辅助以及自身信息利用不足等缺陷,提出基于水体指数与色度分离法的耀斑检测算法以及基于离散余弦变换的耀斑图像恢复算法。首先使用色度分离法提取图像中疑似耀斑的高亮区域,再通过水体指数结合面积阈值和形态学滤波去除散点和孤立区域,进而实现耀斑区域的精准定位和提取。然后依据图像的保真度和局部平滑度构建优化函数,利用离散余弦变换对遥感图像耀斑区域内部的像元进行迭代求解,最终得到恢复后的图像。最后开展现场飞行实验,实验结果表明所提算法能够精确提取和恢复图像水体耀斑区域,并且恢复后的图像在纹理特征和光谱特征两个方面的效果得以改进。
Abstract
Owing to the large errors in the calculation of glint gain in the existing glint recovery algorithms for remote sensing images, the need for near-infrared band information assistance, and the insufficient utilization of remote sensing image information, a glint-detection algorithm based on the water index and chromaticity-separation method and a glint image based on discrete cosine transform are proposed herein. For the proposed recovery algorithm, first, the chromaticity-separation method is used to extract the highlighted areas of the image that are suspected of glint. Then, the scattered points and isolated areas are removed by combining the water index with an area threshold and morphological filtering so as to achieve precise positioning and extraction of the glint area. Then, an optimization function is constructed based on the fidelity and local smoothness of the image, and the discrete cosine transform is used to iteratively solve the pixels inside the glint region of the remote sensing image. Finally, the restored image is obtained. Furthermore, the field flight experiments are conducted. Experimental results show that the proposed algorithm can accurately extract and restore the image water glint area, and the effect of the restored image in terms of texture and spectral characteristics is improved.
李澜, 巩彩兰, 黄华文, 胡勇, 王歆晖, 何志杰, 叶张林. 基于离散余弦变换的无人机耀斑图像恢复算法[J]. 光学学报, 2020, 40(19): 1928001. Lan Li, Cailan Gong, Huawen Huang, Yong Hu, Xinhui Wang, Zhijie He, Zhanglin Ye. Unmanned Aerial Vehicle Glint Image Restoration Algorithm Based on Discrete Cosine Transform[J]. Acta Optica Sinica, 2020, 40(19): 1928001.