Scene Classification Based On Image Perception Hashing

With the widespread use of mobile Internet and the popularization of mobile terminals,digital images have become more and more numerous in the process of generation,delivery,and storage.In the face of these massive amounts of image data,how to organize and manage them effectively become an urgent problem to be solved.The existing method for classifying images is mainly for single-content digital images.Existing methods will not satisfy the effective classification of images when the content in the digital image is complex.Scene classification of digital images is an understanding and analysis of high-level semantics of images.It is based on a variety of characteristics of the image,based on human knowledge and experience of a comprehensive analysis and judgment of the image.The image scene classification method can solve the image classification problem with complex image content.In the process of scene classification,there are two key issues: the semantic gap and the ambiguity of the text description.Image Perception Hashing can solve the above two problems effectively.Image Perception Hashing is a process of mapping digital image information into a digital abstract by a certain method.This mapping method maps the same or similar images to the same or similar abstracts.The different content of images are mapped into different abstracts.This paper proposes a perceptual hashing algorithm based on directional gradient histogram,which has a certain degree of robustness and sensitivity.The scene classification method proposed in this paper is based on the target object in the scene.By analyzing the scene within the target object in the collection process to achieve scene classification.We compare the latent semantic analysis process of the document and use the Probability Latent Semantic Analysis(PLSA)topic model to analyze the subject of the image and then implement scene classification.In the process of implementing scene classification,we first use an image segmentation algorithm to segment the image into different target objects.Then perceive the objects in the image scene by using image-aware hashing.Next,a visual word list based on Image Perception Hashing algorithm is established.Finally,the subject classification model PLSA is used to analyze the theme of each image and classify the image scene.