Digital Art Image and Animation Compositing Technology Based on Computer-aided Drawing Algorithms

Abstract

Addressing the lack of temporal coherence of brushstroke particles, brushstroke misalignment in moving regions, and degradation of stylistic details in digital-art image animation compositing, this paper proposes a digital-art animation compositing method based on optical-flow-guided brushstroke registration and dual-domain adaptive constraints. First, geometric contours and density-transparency attribute fields of brushstrokes are jointly extracted, and rigid temporal constraints are introduced for static regions to suppress texture jitter caused by random sampling. A style-preserving weight parameter is used to adaptively adjust the penalty intensity according to local texture entropy, distinguishing random high-frequency noise from intentional artistic vibration. Second, a multi-scale semantic style encoder is constructed, and global color and texture style consistency are maintained through Gram-matrix statistical constraints. Finally, an optical-flow-guided brushstroke flow field and motion-heatmap attenuation strategy are designed to achieve intelligent following and natural redrawing of brushstrokes under complex non-rigid motion. Experiments on the WikiArt dataset and DAVIS 2023 benchmark show that the method improves temporal coherence, reduces perceptual flicker, lowers endpoint error and FID, and shortens inference time, achieving smooth motion transitions while preserving artistic style.