针对传统基于归一化非完全Beta函数(Normalized Incomplete Beta Function, NIBF)的图像增强方法难以有效地自动获取最优参数且其增强效果受图像动态范围限制的问题, 提出一种基于自适应量子遗传算法的NIBF遥感图像自动增强方法。首先, 由图像色深, 对待增强图像引入最大和最小光谱测度级, 扩大其动态范围。其次, 利用量子比特将NIBF参数编码为量子染色体, 并设置若干量子染色体构成初始参数种群; 对该参数种群进行测量和解码, 以解码值作为NIBF的参数输入, 对图像进行光谱测度变换, 得到对应的增强图像种群。然后, 利用八方向边缘检测模板提取增强图像种群中每个个体的边缘图像, 由边缘强度、边缘数以及熵测度定义刻画参数种群中个体品质的适应度函数, 并以此评价参数种群中的每个参数个体, 保留和记录最优参数个体。在提出的进化策略中, 利用量子旋转门实现量子染色体向最大适应度方向进化, 并根据每代适应度的差异和进化代数自适应地调整量子旋转角的大小; 以最终演化的参数种群中适应度最大的参数个体作为NIBF的最优参数, 生成相应的光谱测度变换曲线, 从而确定输入和输出光谱测度之间的映射关系, 实现图像最优自动增强。对5幅图像的平均实验结果表明: 盲/无参考图像空域质量评价指标提升了122.2%; 自然图像质量评价指标提升了71.8%; 运行时间为10.758 s。满足了遥感图像增强处理中的自动化、鲁棒性和高效率的要求。

Considering the problem that traditional image enhancement methods based on a normalized incomplete beta function (NIBF) have difficultly obtaining optimal parameters automatically and that enhancement effects are limited by the dynamic range of the image, a method of NIBF remote sensing image automatic enhancement based on an adaptive quantum genetic algorithm was proposed. First, from the image color depth, the maximum and minimum spectral measurement levels were introduced into the image to be enhanced to expand its dynamic range. Secondly, the parameters of NIBF were encoded into quantum chromosomes using quantum bits, and several quantum chromosomes were set as the initial parameter population. The parameter population was measured and decoded, the decoded value was input as a parameter of NIBF, and the image was transformed by spectral measure to obtain the corresponding enhanced image population. Then, edge images of each individual in the enhanced image population were extracted using the eight-direction edge detection template. The fitness function of individual quality in the parameter population was defined by edge intensity, edge number, and entropy measure, and each parameter in the parameter population was evaluated and retained, the best parameters of individuals were recorded. In the proposed evolutionary strategy, the quantum rotation gate was used to evolve the quantum chromosomes toward to the direction of maximum fitness level, and the size of the quantum rotation angle was adaptively adjusted according to the difference of each generations fitness and evolutionary algebra. The best parameters of NIBF were the individuals with the most fitness in the finally evolved parameter population, and the corresponding spectral measure transformation curve was generated to determine the mapping relationship between the input and output spectral measure, so optimal automatic enhancement of the image was achieved. The blind/referenceless image spatial quality assessment indicators increase by 122.2%, the natural image quality assessment indicators increased by 71.8%, and the running time is 10.758 s. The proposed algorithm satisfies the requirements of automation, robustness, and high efficiency in remote sensing image enhancement.