Film and photography methods faced numerous challenges in capturing the essence of an image in the past. Initially, apart from the black and white color limitation of images, they faced the difficulty of capturing other elements of the color spectrum, rendering many pictures of famous figures appearing differently than they may have looked. Capturing the identity and pose of the photo’s subject from low-quality antique images is quite challenging.
A new artificial intelligence imaging approach uses color to restyle old photographs such that they have a similar appearance to modern-day photographs. Camera lenses from older days had orthochromatic nature, which incorporated all detected light into the image without discrimination. This resulted in photos appearing grainy and noisy.
Light penetrates the human skin and illuminates the flesh from underneath. In recent years, with the improving computer graphics software, new advanced photographic techniques take advantage of this fact. This illumination helps them eliminate extra noise and wrinkle marks found in many images from the early 1900s.
Time-Travel Rephotography is a technique that uses an archive of contemporary digital portraits to generate sibling photos that share many traits with the colorized black and white photographs. This technique was created by researchers at Google, UC Berkeley, and the University of Washington. It has helped enhance the quality of older photos by adding color and referencing photos taken by modern-day digital cameras to ensure realistic turnout for the appearance of human skin.
Conventional image restoration filters employ independent operations like denoising, colorization, and superresolution. However, the StyleGAN2 framework (used in the Time-Travel Rephotography technique) projects old photos into the space of modern high-resolution images. It achieves all of these effects in a unified framework.
It identifies the shortcoming in quality characteristics of older (black and white) photographs, including graininess and noise, and correcting these issues for the colorized sibling photo modeled after the original. The unified framework achieves deblurring, superresolution, noise removal, contrast adjustment, and colorization, all in one step.
The imaging techniques also allow reflecting on how historical figures may have appeared in their real life. However, the researchers state that more rendering and editing could lead the corrected images of these older photographs to look somewhat different from their original versions.