An image filtering method and system

Abstract

The invention relates to the field of image processing and provides an image filter method. The image filter method includes extracting image characteristics of images; extracting image fingerprints of the image according to the extracted image characteristics; and filtering the images by a bloom filter according to the extracted image fingerprints. The invention further provides an image filtering system. By the image filter method, similar duplicate images identical in main bodies but slightly different in image size, compression quality, chromaticity, contrast ratio and logo can be filtered out, and steps of extracting characteristics of all images and filtering similar images by means of retrieval can be avoided further, efficiency of filtering the similar duplicate images by the bloom filter is higher, and storage and computing resources are low in occupation and consumption.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image filtering method and system. Background technique [0002] With the advent of the era of information explosion, there are a lot of repetitive information on the Internet, and it is a very meaningful subject to effectively identify them. For example, for the crawler system of a search engine, it is meaningless to include repeated pictures, which will only cause a waste of storage and computing resources; at the same time, displaying repeated information is not the best experience for users. For pictures, the reasons for generating duplicate pictures mainly include: [0003] Mirror sites, the URLs of the images are different, but the contents of the images are identical. [0004] The reprint of the image has the same main body of image content, but image processing tools are used to further process the image: such as adding a website logo, scaling the image size, fine-tun...

AI Technical Summary

Problems solved by technology

[0006] However, for the duplicate pictures in the above two cases, they cannot be directly filtered out in the crawler stage, and the duplicate pictures need to be identified and deleted through image processing algorithms.

In the existing near-duplicate image filtering systems, most of the search-based methods are used to search the entire library for the features or fingerprints of the detected images. For example, the patent of Chinese patent application No. 200910146726.5 uses this method. The patent introduces Although the system classifies the image database through the clustering method and reduces the scale of retrieved images, the efficiency of the approximate duplicate image filtering system based on retrieval is relatively low. How to filter images more effectively has become an unsolved problem. question

Images