White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation
Researchers have developed a new framework called VLM-CC to improve cross-camera color constancy in images. This method iteratively refines color balance by using a vision-language model (VLM) to provide feedback on image corrections, rather than directly regressing RGB values. The VLM identifies residual color casts and guides the adjustment process until convergence, achieving state-of-the-art robustness across various datasets. AI
IMPACT Introduces a novel approach using VLM feedback for image color correction, potentially improving visual consistency across different camera types.