Image retrieval method, device and electronic device based on image content

Abstract

Embodiments of the present invention provide an image retrieval method, device, and electronic device based on picture content, wherein the method includes: sequentially performing DCT transformation calculation and binarization processing on the target image, and obtaining the majority of global features of the target image Character string fingerprint information; perform hash coding on the fingerprint information, and group the hash code values ​​in the fingerprint information bit by bit; group each group in the form of key-value according to the fingerprint information value and the hash code value To represent, and based on multiple hash code values ​​in the value value corresponding to each key value, search in different pre-established maps to obtain the retrieval results. The embodiment of the present invention extracts the global feature of the target image, and based on the global feature, uses the combination of hash algorithm and group calculation to perform index search, which can effectively reduce the calculation time while ensuring the accuracy and recall rate. Improve computational efficiency.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of image retrieval, and more specifically, relate to an image retrieval method, device, and electronic device based on image content. Background technique [0002] At present, there are a large number of duplicate pictures in different listings and in the same listing on the platform. On the one hand, it causes a poor experience for C-end users. On the other hand, the same pictures in different listings are suspected to be fake listings. At present, there are traditional image processing methods including SIFT, SURF, and average hash algorithm for the retrieval technology of duplicate images. [0003] The SIFT and SURF algorithms describe the image based on the local features of the image. The hash algorithm based on the mean hash is to sample the global information of the image and generate a hash code, and then use the generated hash code to calculate the distance to evaluate the sim...

AI Technical Summary

Problems solved by technology

[0004] However, among the above two methods, the disadvantage of the former is that the processing speed of the image is slow, and if the image is cropped, it is difficult to judge whether it is repeated, and the detection recall rate of the repeated image is not very high.

The latter still has a certain ability to recognize pictures whose cropping area is less than 25% of the original picture area. However, since the current method of searching for pictures is to directly calculate the distance of the hash code two by two, once the number of pictures is large, the calculation will cost it will be a long time

Images