Image processing method and device and computer equipment

Abstract

The invention provides an image processing method and device and computer equipment. According to the cConvolutional neural network, obtaining feature points corresponding to the training images in the feature embedding space,; and feature center points of the second number of types of images to which the training images belong are obtained, network training is carried out through a central collision mode. Obtaining aA target feature center point corresponding to each type of image is obtained; mapping each obtained feature point and each target feature center point are mapped to a Hamming space; obtaining a first number of Hash codes and a second number of Hash center points are obtained, at the moment, network training can still be carried out in a central collision mode to; obtaining animage hash model, compared with the prior art, the method has the advantages that the similarity between the Hash codes of the training images is learned, the similarity between the feature points ofthe training images and the corresponding central points is learned, the global distribution of the images is learned, the network training time is greatly shortened, and the learning efficiency andthe image Hash code precision are improved.

Description

technical field [0001] The present application relates to the technical field of image retrieval, in particular to an image processing method, device and computer equipment. Background technique [0002] With the increasing number of Internet images, how to quickly and accurately provide users with the required image resources is becoming more and more important. At present, the commonly used image retrieval method is to describe the image content by extracting the underlying characteristics of the image, and then use the feature Compare and judge whether they are similar images. In order to improve the retrieval accuracy, it is often necessary to extract hundreds of thousands of dimensional image features, which requires a huge storage space to save the image feature library, and the workload of each feature comparison is very large. , greatly reducing the retrieval speed. [0003] In order to improve the speed of image retrieval, it is proposed to use hash sequences as im...

AI Technical Summary

Problems solved by technology

[0004] However, in the existing image database processing method based on the hash algorithm, it is often only possible to learn the local distribution of the image. In the era of big data, it is almost impossible to learn the global distribution of the image, which affects the retrieval efficiency. The accuracy of the results, and this method can only get similarity information from a set of compared data each time, the learning efficiency is relatively low, which affects the retrieval efficiency

Images