Abstract:

Semantic scene classification is an important research in the field of image understanding. It is a multidisciplinary subject involving signal processing, pattern recognition, computer vision and cognitive sciences. The image representations and the classification decisions are the two key issues in the scene classification tasks. Image representations try to get the most discriminative descriptions of scene images. Classification decisions try to get a certain computational model different from other categories by learning and training the image samples. Currently, many scene classification methods have been proposed for the researches on the image representations and image classifications. These proposed methods perform the image classification for outdoor manmade scene images, outdoor natural scene images and indoor scene images, and achieve better classification results. However, the changes and differences of the content in the scene image itself can cause both the object differences within the class and the visual similarity between classes, especially for different indoor scene categories. So, it is a challenging problem for the semantic scene classification in the fields of computer vision and cognitive psychology. The research on the scene classification for outdoor images is relatively mature, but the indoor is on the opposite. This paper reviews the researches on the semantic scene classifications, analyzes the performance of the classification algorithms for scene images, and points out the problems of the semantic classifications.