【摘要】 提出一种基于小波的形状和纹理联合特征的图像分类方法。先对图像进行二维小波变换以得到边缘图像,再提取边缘图像的7个边界不变矩组成图像的形状特征向量;在实验中,发现大多数情况下,图像背景的干扰信息大于其对分类的贡献,因此对图像去除其背景,然后在灰度共现矩阵的基础上,计算5个二次统计量作为其纹理特征;最后联合形状和边缘特征向量,并对其进行高斯归一化,用SVM进行分类。结果表明,该方法具有明显的优越性和较强的实用性。更多还原

【Abstract】 This paper presented a method of image classification using wavelet based on shape and texture. Firstly, to use wavelet to transform the images and get edge images, then extract seven edge invariant moments as shape feature vectors. In experiments, finding that the background of images disturb the accuracy of classification in most cases, so to wipe off the background of images and account five second statistics as texture features on the basis of Gray Level Co-occurrence matrix. Lastly, to combine shape features and texture features as image feature vectors, and normalize them by Gauss method, and classify the images by SVM. Experimental results verify the superiority and strong practicability of this method.更多还原