基于图像分块目标区域的检索技术研究

Image retrieval based on block object region

为了解决图像检索中视觉感知缺失所造成的检索率偏低问题,提出了一种基于目标区域检索算法.借鉴方块编码的方法将图像进行二值化处理,从而转化为位平面图,在此基础上,提取图像的背景区域和目标区域,为提取图像的空间分布信息,采用环形分割的方法提取图像的环形分布直方图作为内容特征用于检索.该算法既利用了图像的灰度信息和人眼对纹理的视觉敏感特性,又充分考虑了图像的空间分布对检索效率的影响,试验表明该算法和以往算法相比具有较高的检索效率.

In order to solve the question of the lack of the visual information in image retrieval, an effective and new algorithms is introduced. Firstly, the human visual feature is considered, and the bit-plane image is a- chieved by binary according to the principle of the block truncation code （BTC）. Based on this, the back- ground region and object region is extracted. For introducing the spatial information, the circle distributing histogram is achieved. Not only the color but also significant visual information and spatial information is taken into account. Experiments show that the proposed method has robust retrieval performance.