🤖 AI Summary
To address systematic color bias and high-frequency texture blurring in the DDColor model—caused by insufficient frequency-domain modeling and constrained input dimensionality—this paper proposes a dual-reconstruction framework jointly optimizing color fidelity and frequency representation. Methodologically, we introduce a lightweight color correction module and a learnable frequency compensation mechanism to collaboratively enhance chromatic accuracy and high-frequency detail recovery. Furthermore, we integrate RGB-YUV space mapping, frequency-domain feature disentanglement, and backbone network fine-tuning to enable multi-scale feature co-optimization. Experimental results demonstrate that our approach significantly outperforms the original DDColor model in both PSNR and SSIM metrics. Qualitatively, colorimetric fidelity and structural sharpness are substantially improved, effectively mitigating the model’s inherent chromatic biases and texture degradation.
📝 Abstract
The project has carried out the re-optimization of image coloring in accordance with the existing Autocolorization direction model DDColor. For the experiments on the existing weights of DDColor, we found that it has limitations in some frequency bands and the color cast problem caused by insufficient input dimension. We construct two optimization schemes and combine them, which achieves the performance improvement of indicators such as PSNR and SSIM of the images after DDColor.