Super-resolution reconstruction of multiformat laser range-gated image by the estimate of point-spread function
摘要: 为了解决水下激光距离选通图像成像过程中退化模型复杂的难题,提出了利用连续帧图像估计点扩散函数的距离选通超分辨成像方法。首先,从连续帧图像中选取一帧为参考帧作为初始清晰图像,下一帧图像为模糊图像,用梯度约束的方法求出点扩散函数,用于优化清晰图像;然后,依次将后续帧图像当作模糊图像与清晰图像交替迭代求取点扩散函数并优化更新清晰图像;最后获得的清晰图像与参考帧图像用乘法更新的方法估计点扩散函数,结合凸集投影法算法进行超分辨率成像重建。仿真实验结果表明,改进的算法重建图像分辨率和质量明显优于原始的算法。Abstract: In order to solve the problem of complicated degeneration model of underwater laser range-gated imaging system, a method to estimate point spread function(PSF) by continuous image sequence is presented. At first, a reference image is selected from continuous image sequence as initialized clear image and next sequence image is used as blurred image. PSF can be estimated by constrained optimization using its gradient . The clear image can be optimized and updated by PSF. Then following images are all defined as blurred image to calculate PSF with clear image. Clear image is updated by the PSF. At last, a final clear image can be got from continuous image sequence. PSF of the whole sequence can be estimated by multiplicative updates of reference image and the final clear image. Projection onto convex sets(POCS) can be optimized by the PSF and the super resolution reconstruction can be made. The stimulation results show that the resolution and quality of reconstruction image obtained by improved algorithm are much better than that of orginal ones.
表 1 实验一算法效果对比
Table 1. Comparation with two methods in the experiment Ⅰ
GMG LS · POCS 8.677 4×105 30 075 375 2 215 改进的POCS 9.049 3×105 30 114 613 2 221
表 2 实验二算法效果对比
Table 2. Comparation with two methods in the experiment Ⅱ
GMG LS · POCS 3.456 1×107 221 190 803 1 210.2 改进的POCS 4.211 6×107 273 052 548 1 427.3