基于纹理的图像检索

Image Retrieval based on Texture Featues

为得到纹理特征提取的合适的算法,首先研究了基于灰度共生矩阵的纹理特征的提取方法,将彩色图像变换灰度图像,然后进行四个方向的纹理特征提取,包括能量、熵、惯性矩、相关量四个向量元素作为纹理特征值,并研究了基于Gabor小波的纹理特征的提取。首先将Gabor小波作为母小波,将图像进行二维的Gabor小波变换,将Gabor小波系数的均值和标准方差作为纹理特征值;将两种方法进行比较,查全率和查准率作为测量标准,实验表明基于Gabor小波变换的纹理特征方法在频域具有比较好的检索效果。

To get the right means,the retirval of texture features based on Gray Level Co-occurrence Matrix is reashed.Let color image become gray image,texture features be retrieved form four directions,including energy,entropy,inertia matrix and related vector that are texture features value;Later,the retrival of texture features baseded on Gabor wavelt.First,let gabor wavelt be the mother wavelt,tie image will be transform using the 2-D Gabor wavelt,the average value and standard varianc of the coefficient of wavelt be texture features.Comporised with two means,precision and recall are the measure norm.Retrival of texture features baseded on Gabor wavelt is the better mean in the region of frequence from the experimentation.