CCC: Color Classified Colorization
Mrityunjoy Gain, Avi Deb Raha, Rameswar Debnath
TL;DR
This work tackles automatic colorization of grayscale images with varied object colors by reframing it as a multinomial color-classification problem. It introduces 215 discrete color classes derived from large-scale appearance data and a batch-aware class weighting scheme to balance major and minor colors, supplemented by a DenseNet-based encoder–decoder and a data-driven reduction from 400 to 215 classes. A SAM-based object-selective color harmonization step refines edges and stabilizes minority colors, and the Chromatic Number Ratio (CNR) metric quantifies color richness beyond traditional regression metrics. Empirical results on ADE, CelebA, COCO, Oxford Flowers, and ImageNet show superior CNR and visualization quality, with competitive MSE, PSNR, SSIM, LPIPS, UIQI, and FID scores, demonstrating improved color diversity while preserving overall fidelity. The approach relies on self-supervised training on Place365, enabling robust color distribution learning suitable for real-world images.
Abstract
Automatic colorization of gray images with objects of different colors and sizes is challenging due to inter- and intra-object color variation and the small area of the main objects due to extensive backgrounds. The learning process often favors dominant features, resulting in a biased model. In this paper, we formulate the colorization problem into a multinomial classification problem and then apply a weighted function to classes. We propose a set of formulas to transform color values into color classes and vice versa. Class optimization and balancing feature distribution are the keys for good performance. Observing class appearance on various extremely large-scale real-time images in practice, we propose 215 color classes for our colorization task. During training, we propose a class-weighted function based on true class appearance in each batch to ensure proper color saturation of individual objects. We establish a trade-off between major and minor classes to provide orthodox class prediction by eliminating major classes' dominance over minor classes. As we apply regularization to enhance the stability of the minor class, occasional minor noise may appear at the object's edges. We propose a novel object-selective color harmonization method empowered by the SAM to refine and enhance these edges. We propose a new color image evaluation metric, the Chromatic Number Ratio (CNR), to quantify the richness of color components. We compare our proposed model with state-of-the-art models using five different datasets: ADE, Celeba, COCO, Oxford 102 Flower, and ImageNet, in both qualitative and quantitative approaches. The experimental results show that our proposed model outstrips other models in visualization and CNR measurement criteria while maintaining satisfactory performance in regression (MSE, PSNR), similarity (SSIM, LPIPS, UIQI), and generative criteria (FID).