粒子群蛙跳算法在图像相关反馈中的研究

Image Retrieval Relevance Feedback Algorithm Based on Particle Swarm Optimization-Shuffled Frog Leaping Algorithm

针对在基于内容的图像检索中,底层视觉特征和高层语义特征存在的语义鸿沟问题,提出将粒子群蛙跳算法引入基于内容的图像检索之中.通过使用粒子群蛙跳算法来优化反馈过程,一方面可以增强粒子之间的差异性,使粒子不会停滞于次优解,在检索过程中,检索到更多符合用户需求的图片;另一方面通过粒子种群的迭代寻优,用户对检索到的图片进行评价,可以使计算机更加理解用户的需求.通过仿真实验证明,该方法可以有效地提高图像检索的查准率.

Aim at the semantic gap between visual low level features and high level semantics, this paper proposed a method that imported particle swarm optimization-shuffled frog leaping algorithm into the relevance feedback on the content-based image retrieval. Feedback process is optimized by using particle swarm optimization-shuffled frog leaping algorithm, on the one hand enhance the retrieval ability, makes retrieval can jump out of the sub-optimal , on the other hand through the optimization iteration of the particles that users evaluate the retrieved images, makes the computer understand the needs of users. Through the simulation experiments showing that the proposed method can effectively improve the precision of image retrieval.