一类改进遗传算法的图像信息恢复研究

Study of Image Information Restoration Based on An Improved Genetic Algorithm

首先 ,提出了直接一阶混合映射的灰度转换方程 ,改进了现有的基于远程相关性的图像信息恢复算法 ,既克服了直接映射平滑性差的缺点 ,又避免了一阶线性多项式映射存在的多项式系数分母为零的无意义问题可能带来的重大缺陷 ;接着 ,提出了一类改进的遗传算法作为图像恢复中远端窗口的搜索策略 ,避免了遍历搜索的费时费力 ,又克服了随机搜索的盲目性和不稳定性 ,与基本遗传算法相比加快了收敛进程 ,提高了准确性 .实验结果表明 ,本文提出的两个方面的改进对图像恢复的效果令人满意 。

First, a new class of gray scale transformation equation composed of the direct and 1-order polynomial mixed mapping are proposed, and it improves the existing image information restoration algorithm based on long-range correlation. The improved method not only overcomes the poor smoothness performance caused by the direct mapping, but also solves the possible serious bug brought by 1-order mapping when the denominator of the polynomial equals to zero. Second, a novel improved genetic algorithm used in the search of the remote window in the image restoration is presented. The new algorithm avoids the laboriousness of ergodic search and overcomes the blindness and instability of the random searching. Compared with the simple genetic algorithm, the improved genetic algorithm speeds up the convergent process and improves the accuracy. The experimental results show that the satisfactory performances can be obtained by means of the two improvements above. Especially the novel genetic algorithm demonstrates good robustness in the optimal search of the whole.