提出了一种区域指导的激光水下图像插值算法(简称AIU)。AIU算法首先采用小波变换,对低频成分进行特殊处理,以抑制散斑噪声和增强关注的目标;然后将图像分割成目标区域和背景区域,结合近邻法和众数法确定待插值点所属区域,对区域内部点采用线性插值;对于区域间的过渡点,设计非线性插值公式,对与待插值点属于同一区域的邻域像素,分配较大的权值,对相邻区域的邻域像素分配较小的权值。实验结果表明,AIU算法可以克服传统插值算法的边缘模糊问题,使目标更完整清晰。

An area-directed interpolation algorithm for underwater laser image (AIU) was proposed. Firstly, wavelet transform was performed on the low-frequency components to remove speckle noise and enhance image quality. Then an image was segmented into object area and background area. And whether an interpolated pixel belonged to the object region or not was decided by an approach combining the method of nearest neighbor and statistical mode. The points in the same area are interpolated linearly. For transition pixels between different areas, nonlinear interpolation formulas were designed. Large weights were assigned to neighboring pixels belonging to the same area with the interpolated pixel in the nonlinear formulas, while the neighbering pixels in the different area with the interpolated pixel were assigned with little weights. By choice of different weights, the contrast between object and background can be adjusted to obtain clear object and facilitate object discover and observation. Experiments demonstrate that this method is efficient and feasible.