基于局部区域梯度最小点的遥感图像镶嵌算法

Mosaic algorithm for remote sensing images based on minimum gradient point in local area

寻找最佳拼接缝是遥感图像镶嵌中的重要环节。目前寻找最佳拼接缝的算法大多只考虑到对应像素灰度的差异最小,却忽略了保留目标的完整性。为此,提出一种基于局部区域梯度最小点的最佳拼接缝寻找算法。该算法用梯度表示图像中灰度变化的快慢,梯度值小说明灰度变化较为平缓,据此在图像重叠区域中逐行选取最小梯度值点的连接线作为最佳拼接缝。同时,针对强制改正法易在列方向上产生横向带状效应的问题,提出一种改进的"差值拼接缝消除算法"。实验结果表明,用该文算法确定的最佳拼接缝有效地避开了梯度变化较大的区域,既有效地绕开目标的边缘,又保留了目标的完整性;改进的差值拼接缝消除算法还有效地避免了灰度差较大情况下的横纹现象,使镶崁图像的色调过渡更为平滑,视觉效果得到改善;且算法的运行速度更快,具有一定的可行性和有效性。

Seeking for the optimal seam - line is an important part in remote sensing image mosaic. Most algorithms for finding the optimal seam - line only emphasize the minimum difference between pixels rather than the integrity of objects. Therefore, the authors put forward a new algorithm for finding the optimal seam -line based on minimum gradient point in local area in this paper. This method adopts the gradient to indicate the change of gray scale and then finds the pixel of the minimum gradient in local area progressively according to the principle that the change of gray scale is milder when the gradient is smaller. Meanwhile, an improved difference method to solve the band effect caused by the hard correction method is proposed in this paper. The experimental results show that the optimal seam- line determined in this paper can avoid the great gray scale variation area and the objects can be reserved unbroken. It avoids the band effect well and the seam- line is removed smoothly, the mosaic image also has a good sense of sight. In addition, the proposed method is very simple, effective, and easy to realize.