This paper proposes a novel content-based image retrieval CBIR algorithm model, and also proposes a novel clustering retrieval method suitable for image data, which improves the clustering accuracy by automatically updating the weights of cluster features, thereby improving the retrieval effect. It aims to build a reasonable online teaching model, effectively guide foreign language online teaching in colleges and universities, provide reference for mixed teaching research, and promote students to improve their comprehensive foreign language skills and optimize their learning outcomes. The changes in the grades of the experimental class were also analyzed separately, and the results of the experiment were supplemented by post-study questionnaires and interviews.