To perform colour rendition in digital images for different atmosphere styles is becoming important for effectively communicating visual information. While different camera brands often possess their unique feature styles, precisely reproducing and evaluating colour style effects across diverse camera systems remains significant challenge such as fast mapping and effective evaluation between the source and target styles. To address this issue, a workflow of colour style transfer has been developed. To begin with, multi-device image database including many images was built, comprising 1550 sRGB images, for which each includes two colour charts for calibration, and proposed transfer method using a 3DLUT precisely transfers colour styles, achieving a remarkably low ΔE00 of 1.09 in colour charts tests. Subjective evaluation with 10 volunteers showed a perceptible small visual difference, indicating the effect of workflow achieved satisfactory performance. To overcome the limitations of subjective testing, a Siamese network-based EfficientNet Visual Difference Evaluation Model (EVDM) was introduced, which utilized a lightweight EfficientNet, achieved Pearson correlation coefficients of 0.90 (training), 0.88 (validation), and 0.92 (overall), significantly outperforming sophisticated baseline methods based on CIEDE2000 (max 0.76). This demonstrates EVDM's superior fitting, generalization, and consistency with human perception.