基于熵的图像空间特征提取及检索方法

Spatial feature extraction and image retrieval based on entropy

针对图像颜色的空间分布特征,提出了一种新的基于熵的表示方法。该算法采用空间分布熵来描述颜色的空间特征,具有旋转、平移、尺度等不变特性。结合人类的视觉特征及熵的特性,进一步提出利用加权空间分布熵来优化和改进颜色空间分布特征的提取算法。结合图像颜色直方图,给出了两种图像间的相似性度量方法。仿真实验证明,该算法在进行图像检索时具有更好的检索效果。

According to the color spatial feature of an image, a new algorithm based on entropy is proposed. It introduces color spatial distribution entropy （SDE） as the spatial descriptor of an image, which describes the distribution of pixels with the same color in the image. SDE is also rotation, translation and scale invariant. Integrated with the human vision and the characters of entropy, improvements are made on the new partition method and the weight function. Based on SDE and color histogram, two retrieval methods are proposed to compute the similarity of two images. Experiments show that better results can be achieved than color histogram.