Abstract
<title>Abstract</title> <p>When digital cultural images are integrated into junior middle school art classes, existing applications often lack standardized image preprocessing, computable feature representation, automatic task generation, and data-driven evaluation, making it difficult to connect heritage resources with reusable, repeatable classroom software workflows.To strengthen computer application support for translating cultural images into creative tasks, a modular visual transformation model integrating computer vision, rule-based matching, and intelligent application scoring was developed.Based on the digital Dunhuang murals, the characteristics of line art, color, design, composition and pose were extracted. Course tasks were generated with feature task matching, difficulty classification, and student work coding, and 160 students were tested for six weeks in four classes.The results showed that the experimental group showed more significant improvements in the extraction of visual elements, formal reorganization, and thematic expression. The correlation between AI and teacher scores was 0.86, and the mean absolute error was 3.84. The results show that this application framework improves digital heritage processing, instructional generation, evaluation, and iterative optimization. CCS Concepts Applied computing → Arts and humanities → Fine arts</p>