一种基于加权颜色形状特征和LBP-GLCM纹理特征提取的服装图像检索方法

A Clothing Image Retrieval Method Based on Weighted Color Shape Feature and LBP-GLCM Texture Feature Extraction

为了提高颜色和纹理特征在服装图像检索中的准确性,本研究提出一种基于加权颜色形状特征和LBPGLCM纹理特征提取的服装图像检索方法以提高服装图像检索的精度。首先,在第一道检索中运用颜色直方图和Hu不变矩的特性进行加权处理,而后,再运用Uniform-旋转不变局部二值模式算子对返回的结果进行处理,得到相对稳定的编码图像。实验证明,此方法能将不同特征的优势进行融合,使之相互补充,检索结果较之单一特征的检索更加准确。

In order to improve the accuracy of color and texture features in garment image retrieval, a garment image retrieval method based on weighted color shape feature and LBP-GLCM texture feature extraction is proposed to improve the accuracy of garment image retrieval.Firstly, the color histogram and Hu invariant moments are used in the first retrieval to process the weights. Then, the Uniform-rotation invariant local binary pattern operator is used to process the returned results, and a relatively stable encoding image is obtained.Experiments show that this method can fuse the advantages of different features and make them complement each other. The retrieval result is more accurate than that of single feature.