基于拉普拉斯特征映射法的水下图像降维研究

Research on underwater image dimensionality reduction based on Laplacian Eigenmap

由于水对光有强烈的衰减作用,使得水下彩色图像许多的重要信息丢失。为了从水下彩色图像的低维结构中获取更多细节信息,采用一种局部非线性的拉普拉斯特征映射算法,对水下彩色图像进行降维处理,由高维到低维的特征映射得到二维嵌入的结果,应用改进的重投影方法获得重构的图像。从低维结构图结果中可获得水下彩色图像没有体现出来的一些细节,提高了图像的对比度,并且可以观察出光在水下的分布规律,有助于对水下成像进一步的研究。

Since water has a strong attenuation effect on light,much more important information in the underwater colour images is lost.In order to obtain much more detailed information from the low dimensional structure of underwater images,a lo cal nonlinear Laplacian Eigenmap algorithm is used for the dimensionality reduction processing of underwater color images,a two dimensional embedded image is obtained from high dimensional to low dimensional feature map,and a new reprojection method is asopted to gain a reconstructed image.Some image details that the underwater color image does not reflect can be obtained from the low dimensional image.The contrast of images was improved and the light distribution pattern in the underwater could be observed by the algorithm.It helps to further study of underwater imaging.