一种基于联合表示的图像分类方法

在图像分类识别中，对于同一目标的不同图像，其训练样本和测试样本在同一位置的像素强度通常不同，这不利于提取目标图像的显著特征。这里给出一种基于稀疏表示的联合表示的图像分类方法，此方法首先利用相邻列之间的关系得到原始图像对应的虚拟图像，利用虚拟图像提高图像中中等强度像素的作用，降低过大或过小强度像素对图像分类的影响；然后用同一个目标的原始图像和虚拟图像一起表示目标，得到目标图像的联合表示；最后利用联合表示方法对目标分类。针对不同目标图像库的实验研究表明，给出的联合方法优于利用单一图像进行分类的方法，而且本方法能联合不同的表示方法来提高图像分类正确率。

Syncretic representation method for image classification

In the classified recognition of image, for different images of the same object, because the pixel intensities of the training samples and the test samples at the same positions are usually different, it brings difficulty to extracting the salient features of the object images. This paper proposed an image classification method based on syncretic representation of the sparse representation. Firstly, the virtual image of an original image is obtained by using the connection between adjacent columns of the original image, the virtual image is utilized to enhance the importance of the pixel with moderate intensity and reduce the effects of the pixels with overlarge or undersize intensity on image classification; then the original image and virtual image of the same object are together used to represent the object, so as to obtain the syncretic representation of an object image; finally, the syncretic representation method is used for object classification. The experiments on different object image libraries show that, the given syncretic method is superior to the classification method realized by utilizing single image, in addition, by combining the method with other different representation methods, the accuracy of image classification can be improved.