【摘要】 图像主题区域是表达图像主要语义内容的部分.根据图像不同区域的视觉显著性差异特点,利用改进的视觉注意模型,提取图像的主题区域,并划分为不同语义类别.提出基于主题区域的图像检索方法,该方法能按照图像主题区域所属的语义类别,在图像之间有针对性地选择同类别主题区域进行区域相似性比较.实验结果表明,此方法可以更有效地表达图像的主要语义内容,提高图像检索的效率和准确性.更多还原

【Abstract】 The image thematic regions describe the key semantic contents in an image.In this paper,we presented a method of the automatic extraction of the image thematic regions.Use visual attention computational model to extract the salient regions that lie in the image.Analyze how the salient and non-salient regions express the image themes based on the properties of size,position and distribution.Extract the thematic object regions and thematic background regions automatically.Introduce the thematic region-based image retrieval.This method can efficiently choose different image thematic regions to match between images based on the representation of the image themes.Results in our experimentation demonstrate the method introduced in this paper can efficiently describe the key thematic contents of images,increase the efficiency and accuracy of image retrieval.更多还原