Optimization of cross-cultural communication model for ethnic minorities based on self-similarity and comparative learning

Abstract

The super-resolution algorithm of self-similarity is utilized in this paper to construct an image training set based on the multi-scale self-similarity of images and reconstruct the super-resolution of images. The visual question-and-answer method of contrast learning ensures full coverage of key targets, which makes the optimization of mutual information more reliable and stable to construct a cross-cultural communication model for ethnic minorities. The results show that compared with the cross-modal audio-video instance discrimination model, the accuracy of TOP1 at the visual clip level is 3.04% higher, and the accuracy of TOP5 at the video level is 2.62% higher for the model designed in this paper. This paper's design model can enhance the ability of cross-cultural communication among ethnic minorities, as indicated.