利用位平面欧拉-熵向量的图像检索方法

Bit-plane Euler-entropy vector based for image retrieval

首先将图像分解为8个位平面,选择前4个重要位平面,求出其灰度码表示,计算出每个灰度码位平面的欧拉数,组成欧拉向量;其次计算出每个灰度码位平面的信息熵,组成位平面熵向量;再把欧拉向量和位平面熵向量特征综合为一个新的图像特征,最后采用马氏距离计算图像间的相似度.实验结果表明本方法维数低,速度快,且避免了图像量化造成的误检,与比位平面熵算法和欧拉向量算法相比,检索效率高.

Firstly, the image was divided into 8 bit planes. Considering the first four most significant bit planes of the image and liter converting each 4 bit vector to its corresponding reflected gray code, we computed Euler number of each bit plane to build the Euler vector. Secondly, an entropy vector was constructed by computing the entropy of each gray-code bit plane. Then, the euler vector and entropy vector were combined to make a new image feature. Finally, Mahalanobis distance was adopted to measure the similarity between images. Experimental results are reported to demonstrate that the method was fast, low dimension and no improper retrieval caused by image eolor quantization technique, and the method had the better performance than Euler vector method and bit-plane entropy method.