An image search method and system based on focus object recognition and topic semantics

Abstract

An image search method and system based on focus object recognition and theme semantics is provided. The method comprises the following steps: S1, the low-level feature semantics of a to-be-searched image is extracted, the image is segmented after the formation of a high-dimensional feature vector space to acquire elements of the to-be-searched image; S2, the elements of the to-be-searched image are recognized and abstracted to a semantic concept layer, whether the different elements in the to-be-searched image are related is judged according to the relative distance and the co-occurrence probability of the different elements, and if so, the related elements are combined into an object; S3, the relative space position and the specific area of the object are calculated, the interest index and the important index of each object are also calculated, the focus object is then recognized from the different objects according to the calculation result; and S4, a theme semantics model for the to-be-searched image is established, the existing image with the focus object is acquired, then the similarity between the to-be-searched image and the existing image is measured through the KL distance, and the existing image with the higher similarity is outputted.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an image search method and system based on focus object recognition and theme semantics. Background technique [0002] With the rise and vigorous development of the mobile Internet, all kinds of smart terminal devices are rapidly popularized. Mobile developers have launched a large number of applications (apps) in mobile scenarios, such as client-side social application software such as WeChat, Laibian, SnapChat, and Instagram, and client-side e-commerce software such as Taobao, Tmall, and JD.com. Such software will not only generate a large amount of image data with rich content, but also generate various image search needs. Especially with the development of e-commerce, the rapid layout of the O2O (Online To Offline) model, and the emerging mobile search form of "phone camera + image search + price comparison shopping" pose a huge challenge to traditional ...

AI Technical Summary

Problems solved by technology

However, this method has the following problems: it is impossible to manually label a large amount of image data, and it is highly subjective; when the image content is rich, it is difficult to accurately describe it in words

From the analysis of the results returned by the two image search engines, it can be seen that: Baidu’s search results reflect a certain amount of fuzzy matching, but they are not fuzzy enough, resulting in poor scalability. feature semantics (color, texture, shape), and did not return pages containing an exact match for the image; while Google's search results, although containing both pages that exactly match the image, also contain similar-looking images, but these similar-looking images Irrelevant pictures such as "black USB flash drive, black notebook, black telescope" appeared in the website, which reflected Google's fuzzy matching strategy was too fuzzy and the matching was inaccurate

Therefore, the image retrieval system currently representing the highest level in the industry (Baidu image search engine, Google image search engine), the performance of similar image search is not satisfactory to users

Images