基于混合搜索算法的图像稀疏分解

Image sparse decomposition based on hybrid search algorithm

为了提高图像稀疏分解的效果,降低其计算时间,提出一种基于量子进化算法(quantum-inspired evolutionary algorithm,QIEA)和改进差分进化算法(improved differential evolution,IDE)的混合搜索算法,并应用到图像稀疏分解中。该方法将IDE引入到QIEA中,前期进行QIEA寻优,当寻优搜索到的最优解经过多次进化后没有变化时,引入IDE以提高搜索解的精度和质量。图像稀疏分解的仿真实验结果表明,与QIEA和IDE相比,混合搜索算法的图像稀疏分解方法获得的重构图像具有最好的图像视觉质量和最高的峰值信噪比,且具有相对较低的计算时间。

To improved the results for image sparse decomposition and reduced the computing time,a hybrid search algorithm（HQIDA）based on quantum-inspired evolutionary algorithm（QIEA）and improved differential evolution（IDE）is proposed,and applied to image sparse decomposition.This approach introduces IDE into QIEA for searching best solution.In the beginning,it uses QIEA to search optimal solutions,if the searched optimal solution is not improved in certain successive generations,IDE is applied to improve the accuracy and quality of the optimal solution.The performance of HQIDA is tested on image sparse decomposition.Experimental results show that HQIDA is superior to QIEA and IDE in terms of the quality of solutions,its reconstructed image has the best visual quality,the highest peak signal to noise ratio,and relatively shorter computing time.