基于教与学优化算法的相关反馈图像检索

Relevance Feedback Image Retrieval Based on Teaching-Learning-Based Optimization Algorithm

为提高基于内容的图像检索的检索性能和检索速度,克服低层视觉特征与高层语义概念间的"语义鸿沟",提出一种基于教与学优化的图像检索相关反馈算法(TLBO-RF).结合图像检索问题的特殊性和粒子群优化算法的优点,对TLBO算法中个体的更新机制进行了改进,通过将相关图像集的中心作为教师以及引入学员最好学习状态Pbest,使之朝用户感兴趣的相关图像区域快速收敛.将该算法与目前效果最好的两种基于进化算法的相关反馈技术在两套标准图像测试集上进行对比,结果表明本文算法相较于另外两种算法具有明显的优势,不仅提高了图像检索性能,同时也加快了图像检索速度,更好地满足了用户的检索要求.

To improve the performance of image retrieval,and accelerate the speed of image retrieval in content-based image retrieval and reduce the＂semantic gap＂between visual low-level features and high-level semantic,relevance feedback image retrieval based on teaching-learning-based optimization algorithm is proposed（ TLBO-RF）. Considering the specificity of image retrieval and the advantage of the PSO,the update strategy of individual is modified in TLBO,the center of the relevant images is regarded as the teacher and the personal best is introduced,which makes the algorithm converge fast to the region of relevant images that the user is interested in. TLBO-RF is compared to two state-of-the-art RFs based on evolutionary algorithm on two benchmark images. The results showthat TLBO-RF has obvious advantage in comparison with other two algorithms,not only increases the performance of image retrieval,but also improves the image retrieval speed,and can better meet the user needs of image retrieval.