结合彩色边缘与LBP纹理的图像检索算法

Image Retrieval Based on Color Edge and LBP Texture

基于图像颜色、纹理和形状单一特征的特征提取和匹配的方法很多,各有优缺点,因此,本文提出了将图像中心区域的uniform模式的LBP纹理与环形分块彩色边缘相结合的图像检索算法。常用的形状检索算法只能对连续封闭曲线才有好的检索效果,而对自然彩色图像检索效果较差,而本文将图像分成几个环形分块,对每一环形分块内的图像提取彩色边缘并形成颜色直方图用于图像形状描述。纹理采用uniform模式的LBP描述,最后采用加权法融合形状特征和纹理特征。根据实验比较,该算法能较大提高大多数类别图像检索的查准率。

Feature extraction and matching method based on single feature such as image color, texture and shape are many, each has advantage but also has some shortcomings, thus puts forward the uniform model of the central region of the image of LBP texture combined with a circular block color edge of image retrieval algorithm. The shape of the commonly used retrieval algorithm can only has a good retrieval effect of continuous closed curve, and the natural color image retrieval effect is poorer. This article separates the image into several annular blocks, and then uses form of color histogram to image the shape description for each ring image edge extraction of color, and uses the uniform model of LBP to describe texture. Finally, shape features and texture features are incorporated by using weighted method. According to the experimental comparison, our method can improve greatly and most of the categories of the image retrieval precision.