改进的双链量子遗传算法在图像去噪中的应用

Improved quantum genetic algorithm with double chains in image denoising

摘要:

针对传统双链量子遗传算法收敛速度慢、搜索精度低、鲁棒性差等不足，提出一种F型双链量子遗传算法（F_DCQ-GA）.对编码空间进行单值映射处理，在保证量子种群适应度值与相应幅角排序单调性的前提下，缩小算法的搜索空间，增加搜索密度；在量子更新时引入自适应步长因子，使步长随目标函数在搜索点处梯度的变化而变化，有效解决了传统寻优算法普遍存在的全局最优解搜索困难的问题；在染色体变异更新时提出了π/6门，克服了原来非门变异无法更新量子比特概率幅的缺点.将F_DCQGA优化算法应用于小波阈值去噪的阈值选择机制中，通过仿真证明F_DCQGA优化算法提高了小波阈值函数的收敛速度和搜索精度，在图像边缘特征提取中可以获得更小的均方误差（S_ME）和更大的峰值信噪比（R_PSN），同时又保留了大部分高频信息.

Abstract:

To solve the problems of slow convergence speed, low search precision and poor robustness in traditional double chains quantum genetic algorithm, a new double chains quantum genetic algorithm (F _DCQGA) is proposed. The coding space is mapped to reduce the algorithm searching space and increases searching density, under the premise of guaranteeing quantum population adaptation and argument population monotonicity. The adaptive step-length factor is introduced to the quantum updating, which changes the step-length with gradient of objective function in searching points. This could solve the global optimal solution search difficulties caused by oscillatory occurrence in traditional optimization algorithm. Quantum π/6 gate is presented in chromosome mutation upadating, to overcome the shortcoming that NOT gate can not update quantum bit probability amplitude. The F_DCQGA is applied to the threshold selection of wavelet threshold denoising. Simulation results show that F_DCQGA improves the convergence speed of the wavelet threshold function and searching precision. And in image edge feature extraction, the smaller mean square error (S_ME) and larger peak signal to noise ratio (R_PSN) are gained. Simultaneously, the high frequency information is also retained.