colorization

Colorization is the process of adding color to grayscale images or videos. The technique can be applied to still photographs, motion picture footage, or other monochrome media. The objective may be to approximate real-world colors, to produce a particular artistic effect, or to aid interpretation when color information has been lost.

Historically, colorization began with manual painting on film frames and later shifted to digital methods. Early digital colorization in the late 20th century used frame-by-frame or semi-automatic processes. In film restoration, colorization has been controversial because it modifies original material and can alter historical perception, prompting discussions about authenticity, consent, and the artist's intent.

Modern colorization employs a spectrum of approaches. Manual colorization relies on artists painting color into grayscale images, often guided by reference material. Automated and semi-automatic methods use priors, such as spatial coherence and learned color distributions. Deep learning approaches train neural networks on large color image datasets to predict plausible colors from grayscale input. Some systems allow user-provided hints or reference color images to guide the result, while others operate in an end-to-end fashion.

Applications include film restoration, archival enhancement, visual effects, and research into perception and history. Because colorization can introduce biased or anachronistic colors, practitioners emphasize documenting the method and limitations, and often label colorized material as colorized rather than shot in color. The accuracy of color is typically probabilistic rather than guaranteed.

Colorization continues to evolve with advances in machine learning and image processing, balancing fidelity to the original material with creative or practical objectives.