基于颜色特征和SVM的自然图像分类标注算法

Natural Image Automatic Annotation Algorithm Based on Color Feature and SVM

针对传统的基于颜色特征图像分类算法运算复杂度高和图像误匹配等问题,提出了一种基于支持向量机的自然图像分类标注算法.该算法通过Matlab编程实践,在彩色空间量化中引入图像分块理论以解决全局直方图中空间位置信息缺失和色彩数量庞大等问题,再通过支持向量机筛选出与待处理图像颜色特征相似的样本子库,以提高图像重建质量,保证图像分类精度.实验结果表明,该算法由于加大了中心位置图像子块的权重,能够更好地提取关键特征信息,保证了图像匹配精度,能比较准确地对图像进行分类标注.

Directing at the problems such as high computational complexity and image mismatching of traditional image classification that based on color feature,this article presents a natural image automatic annotation algorithm that based on SVM. Through Matlab,this algorithm introduces the theory of image block into color spatial quantization to solve the problems of information missing of spatial position and huge quantity of colors in global histogram. In order to improve the quality of image reconstruction and to ensure the accuracy of image classification,the SVM is used to screen the sample sub-library which is similar to the color feature of the processing image. The experimental results show that by increasing the image sub-block in center,this algorithm is able to extract the key features of information better and ensure the accuracy of image matching,so that it can more accurately classify image annotation.