结合图像熵和分块匹配的图像检索

Image Retrieval based on image entropy and blocked matching

针对图像颜色的空间分布特征,首先计算出图像整体熵并且与设定的阈值进行对比,根据相似的图像信息熵也相近来筛选出整体相似图像,实现对图像的预分类。然后再对所得图像不固定分块,分别计算以灰度共生矩阵表示的纹理特征,进行更深层次的检索。最后利用欧几里得矩来进行相似性度量。仿真结果表明,本文提出的方法检索图像更精确、更迅速。

Considering the space distribution feature of color, first we calculate image＇s overall entropy and make a contrast with set threshold. Based on the theory that similar images have similar image information threshold, we select images similar in general, so pre -classification is realized. Then block irregularly as to these selected images, and work out their gray co -occurrence matrix as their veins feature respectively to do further processing. Then we measure similarity by Euclidean Distance. Experiment shows the way this paper forwarded make the image processing more exact and more quick.