参数优化的PCNN图像检索

Parameter Optimization in PCNN for Image Retrieval

脉冲耦合神经网络(PCNN)用于图像检索时需人工确定较多参数,参数确定的好坏严重影响检索效果,针对以上问题,提出一种基于进化学习的参数优化方法。通过引入粒子群算法(PSO),构建优化目标函数,提前对图像库中少量图像进行分类训练,对脉冲耦合神经网络的各参数进行优化并用于图像检索。实验表明,提出的算法能有效找到各参数的近似最优解。对图像库中未训练图像进行检索时也取得较好效果,在检索查准率、查全率及主观视觉效果方面本文方法均优于经验参数。

When PCNN was used for image retrieval, the manual parameters selection became the difficulties and whether the selected parameters are good or not determines the retrieval results mostly. A novel method based on evolutionary learning for optimizing the parameters of Pulse Coupled Neural Network (PCNN) was proposed to overcome these problems. Firstly we classified some images of the database and trained them in advance by introducing the Particle Swarm Optimization (PSO) and restructured the fitness function to optimize the parameters which were used in image retrieval. Experimental results show that the proposed method can achieve the optimal parameters adaptively, and the retrieval results perform well even in the untrained images. The retrieval results convince that the proposed method was better than experienced parameters on precision ratio, recall ratio and personal visual judgment.