An image collection abstract generation method in social media

Abstract

The invention discloses a method for generating abstracts of image collections under social media. Firstly, visual features of images in image collections under social media are extracted; then the extracted local features and global features are fused and matched; The spatial constraint relationship forms a spatial location map and a simplified geometric constraint representative data set to further filter out noise points; secondly, an automatic clustering model based on neighbor propagation is established; finally, a group of optimal clustering centers is found by using the clustering validity index. The present invention can make full use of the multi-modal features of images, and summarize a large number of related images in image collections under social media, so as to accurately generate a group of optimal representative images.

Description

technical field [0001] The invention relates to the field of analysis and retrieval of social media image collections, in particular to a method for generating abstracts of image collections under social media. Background technique [0002] With the rapid development of Internet technology and the wide application of multimedia technology, the status of the Internet has been continuously improved, and it has become the main platform for people to communicate and share information. Network users spontaneously contribute multimedia materials such as pictures and videos, and disseminate them through sharing, evaluation, discussion, etc., so that a large number of pictures flood into the network. In such a vast ocean of pictures, it is not easy to find the one we want. When we initiate a query, what is returned to us is a huge queue of multimedia picture sets, and there is a large amount of data with repeated or partially repeated content, as well as a considerable amount of da...

AI Technical Summary

Problems solved by technology

The smaller the number of clusters, the simpler the calculation but the lower the accuracy of the data set division. The higher the number of clusters, the higher the accuracy of the data set division but the redundancy of the summary image.

[0006] Second, traditional clustering algorithms rely on initial cluster centers, so the results are often unstable

[0007] Third, an optimal clustering algorithm does not exist, and no clustering method is suitable for all data sets

Images