Photos from history are no longer bound to black and white. Through groundbreaking editing software conceived by a team of researchers at UC Berkeley, photographers will able to spruce up black and white images by infusing them with accurate color schemes according to perceptions of the human eye. Otherwise known as a “convolutional neural network,” this smart technology infers the most probable colors of black and white photos, and then alters them with new pigments so that they appear realistic.
“Computer vision algorithms often work well on some images, but fail on others,” Richard Zhang, Phillip Isola and Alexei A. Efros write in their report. “We believe our work is a significant step forward in solving the colorization problem.”
Thanks to the plethora of new technical capabilities available today, photographers and producers might normally go out of their way to revert color images to grayscale ones. Interestingly, though, the belief is that this reverse editing software will liberate the process, allowing photos to be rebuilt with color based on a data set of more than one million color images.
“Previous approaches have either relied on significant user interaction or resulted in desaturized colorizations,” the researchers said in describing how their automatized technology will reduce imprecise color assignments. "Our method successfully fools humans 20 percent of the time.”
They add that this figure is “significantly higher than previous methods.”
Curious about how some of your grayscale photos might look in color? Try the software for yourself, although it may require some coding skills since it has not yet been integrated into editing programs.
Either way, let us know what you think: Will a more accurate color-injection process totally disrupt photography as we know it? Tell us your thoughts in the comments below!