Based on Diffusion Models for Learning Painting Color Distributions and Digital Image Generation Methods
Abstract
Diffusion-based generators have rapidly become the dominant paradigm for high-fidelity image synthesis, yet most existing systems still optimize semantic realism more strongly than chromatic structure. This mismatch is especially problematic for painting-oriented generation, where palette organization, inter-hue balance, and style-consistent color density are central to visual quality. This paper proposes a palette-aware latent diffusion framework for learning painting color distributions and generating digitally synthesized images with improved chromatic fidelity. The method encodes reference palettes as quantized color tokens in CIELAB space and injects them into the denoising backbone through cross-attention and cross-scale chromatic transport. A color-distribution loss is further introduced to align the generated histogram with the target palette manifold while preserving semantic composition in latent space. To improve stability, the model jointly optimizes noise prediction, palette anchoring, and reconstruction consistency. Experiments on WikiArt and zero-shot transfer to MS-COCO prompts show that the proposed framework reduces FID from 15.9 to 12.6 relative to a strong SDXL-based baseline, lowers color earth mover's distance from 0.059 to 0.041, and improves average CIELAB ΔE from 9.1 to 6.7. Qualitative results indicate that the model better preserves intended warm-cool balance, painterly harmony, and cross-image style consistency. The study demonstrates that…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.