基于HMM的自动图像标注方法

摘要

Automatic image annotation is a feasible way to reduce "semantic gap". The hidden Markov model (HMM)-based automatic image annotation method proposed in this paper can effectively mine the semantic visual feature distribution of each keyword in order to set up the corresponding relations between the image and the keyword; moreover, it can also efficiently attain the semantic correlation by combining the co-occurrence relation of the keywords. Therefore to build correlated multi-prospective models of image-keyword relation and keyword-keyword relation will be helpful to fulfilling the task of automatic image annotation. At last, the validity of the proposed method is verified by a series of experimental results conducted on COREL image database.%自动图像标注技术已经成为弥补"语义鸿沟"的一种有效途径.提出基于隐马尔科夫模型HMM(Hidden Markov Model)的自动图像标注方法,不仅有效地挖掘关键词的语义视觉特征分布,从而建立图像-关键词的对应关系;而且通过融合关键词的共生关系,高效地获取关键词-关键词的语义关联.为此,建立图像-关键词与关键词-关键词的多视角相关模型,有助于解决自动图像标注任务.最后,在COREL图像数据集上的一系列实验结果,验证了提出方法的有效性.