The invention requests to protect an image description method based on a 2D local binary pattern (2D LBP), and relates to the technical fields of digital image processing and machine vision. The method comprises the following steps: first, preprocessing an image in a data set to make the image change from a color image to a gray image, and using a rotation invariant uniform local binary pattern method (LBPriu) to extract the features of the preprocessed image to obtain a local binary pattern diagram (LBP diagram) of the image; then, determining the window size of the LBP diagram, constructing 2D LBP features on the LBP diagram, changing the radius size of an LBPriu neighborhood, and fusing the 2D LBP features corresponding to different radiuses to construct a final feature for describing the image; and finally, selecting a training set and a testing set, and matching the training set and the testing set to obtain a classification result. The context information of the LBP value of the image is utilized, and the accuracy of image description is effectively improved.