Skip to Main Content
In this paper, we will propose a new approach about how to categorize viewfinder images captured by digital cameras. This new unsupervised learning approach based-on visual characters of the images and LDA (Latent Dirichlet Allocation) model. In this approach, we represent the image of a scene by a collection of local regions, denoted as codewords. The image codewords which include visual characters of images are training units prepared for image categorizing. After learning, we get the LDA model parameters for each category image, and then in categorizing, we classify the input image by selecting the best category model. The test result shows that this approach can categorize different scenes automatically and works well.