An Image Classification Method Based on Matching Similarity and TF-IDF Value of Region

Traditional image classification methods are mainly based on the overall color statistics and content semantics of the image itself. However, due to the poor distinctiveness of overall color statistics and content semantics, traditional methods often cannot acquire very accurate classification results. In this study, an image classification method based on matching similarity and TF-IDF value of region is proposed. First, an image is divided into some fixed-size regions according to its size, and features of each region are extracted and stored. Then, the matching similarity and TF-IDF value of each region are obtained by comparing with and matching regions in the standard image region dataset. Finally, constructing corresponding feature vectors based on features of each region to get the final image classification result. Experimental results show that the proposed method can be used for fast classification of vast images and gains good classification results.