Semantic annotation method, device and client for image

Abstract

The invention provides a semantic annotation method, a semantic annotation device and a client for an image. The method comprises the following steps: extracting word information in an image; generating annotation information according to the word information; annotating the image according to the annotation information. According to the method provided according to the embodiment of the invention, on one hand, a process for manually screening image category training classifiers in a traditional semantic annotation method of the image can be omitted, the labor and the time are saved, a semantic gap between the low-level feature and semantic information of the image is avoided, and the stability and coherence of the semantic annotation of the image are improved; on the other hand, an inherent problem that semantic tags are limited can be overcome, the completeness of a semantic annotator of the image is improved, contents in the image can be accurately described by semantic annotation information, and meanwhile, the semantic annotation speed of the image is increased.

Description

Technical field [0001] The present invention relates to the field of computer technology, and in particular to a method, device and client for image semantic annotation. Background technique [0002] Automatic image semantic annotation refers to the process of using computer vision technology to allow the computer to understand the content of the image, thereby automatically marking the image with text semantics. In other words, automatic image semantic annotation is a process for the computer to accurately describe and understand the semantics of the image, and its essence is to establish the relationship between the image and the text semantics. Image semantic automatic annotation is usually used in different scenarios such as image retrieval, image classification and album management. [0003] At present, automatic image semantic annotation methods usually use pattern classification to establish the association between image and text semantics. Specifically, each classificatio...

AI Technical Summary

Problems solved by technology

[0004] At present, the existing automatic semantic annotation methods for images have the following three problems: 1) The existing automatic semantic annotation methods for images solve the semantic annotation of images as a classification problem, and usually require manual selection and annotation of semantic categories , to train the semantic classifier

Therefore, manual labeling of images for each category is time-consuming and laborious, and it cannot exhaust the entire semantic space, resulting in a very incomplete semantic description of the image; 2), each semantic category corresponds to a semantic classifier, and the semantic When labeling, all semantic classifiers need to be tested, and the semantic category with high confidence is selected as the semantic label of the image

However, if the number of semantic categories is huge, the amount of computation for semantically annotating images will also be huge, so that the existing methods for automatic semantic labeling of images cannot be applied in scenarios with high real-time requirements; 3). There is a semantic gap between the low-level features of the image and the semantic information. Therefore, the existing methods for automatic semantic annotation of images have poor stability and consistency when acquiring semantic categories.

Images