Leveraging Multi-Model Learning for Enhanced Grayscale Image Colorization

This study addresses the challenge of grayscale image colorization, which often suffers from inconsistent results and an inability to manage complex patterns in single-model approaches. A multi-model ensemble approach is presented for grayscale image colorization, which integrates five advanced deep learning models. These five models were selected rather than two or three to increase the robustness across multiple datasets and to capture a greater range of features. By combining the results from models trained in both RGB and Lab color spaces, the proposed method improves visual quality and color accuracy. The study demonstrated that the ensemble approach is much better than a single-model approach when considering the colorization accuracy and the fine texture. Colorization plays a vital role in preserving historical photo restoration, enriching digital archives, and improving visualization across education, science, and the arts. By transforming grayscale images into accurate color representations, it also paves the way for future advancements in digital storage and media restoration.