Method and system for generating an image search engine

Abstract

The invention provides an image search engine generation method and system. The system comprises a vision word generation module, a vision word correction module and a search engine generation module, wherein the vision word generation module is used for determining vision word sets of to-be-indexed images based on a vision word bank; according to provider information of the to-be-indexed images, the vision word correction module sets the to-be-indexed images provided by each provider into a group for the provider, vision word statistics and judgment are conducted within the range of the to-be-indexed images in each group, unavailable vision words relevant to the corresponding provider are determined, vision word sets of the to-be-indexed images in the group are corrected respectively, and the unavailable vision words relevant to the corresponding provider are removed from the vision word sets; the search engine generation module is used for adding the corrected vision word sets of the to-be-indexed images to a reversely-arranged index of an image search engine in the image search engine generation process. The image search engine generated through the scheme can effectively avoid the influences on a search by non-main regions of the images.

Description

technical field [0001] The present application relates to image search, and more specifically, to a method and system for generating an image search engine. Background technique [0002] The image search engine based on local features can find similar images in a massive image collection based on a given image, and is currently the mainstream content-based image search engine. [0003] The frameworks of image search engines and text search engines are generally similar, but one of the most important differences is that "text words" are replaced by "visual words". In an image search engine based on local features, it is necessary to convert the image into a list of "visual words", and then build an inverted index (inverted index) for the image according to the visual words. [0004] In recent years, image search engines based on local features have gradually entered into commercial applications. There are general fields and some subdivided fields. One of the important subdiv...

AI Technical Summary

Problems solved by technology

However, this solution cannot solve the problem that the subject is partially blocked by the logo, and at the same time, the existing product subject area recognition cannot achieve a sufficiently accurate effect

[0008] For images provided by other providers, if the non-subject areas of the image are the same or similar, the existing methods cannot effectively remove the impact of these non-subject areas on the search

Images