Project

Colorchives

 Nina Lutz

Colorchives represents a continuous and interweaving research thread by Nina Lutz.

Color quantization is a long studied issue in computer graphics. The idea of color quantization is utilizing decompression to represent an image in less pixels than its original format. Color quantization is also utilized to generate color palettes and information from images by converting colors into geometric color spaces, represented below.

Colorchives represents a continuous and interweaving research thread by Nina Lutz.

Color quantization is a long studied issue in computer graphics. The idea of color quantization is utilizing decompression to represent an image in less pixels than its original format. Color quantization is also utilized to generate color palettes and information from images by converting colors into geometric color spaces, represented below.

The original goal of Colorchives was to develop a prototyping tool for designers to generate color palettes, specifically towards event-based photos and photos with human subjects for archival and representational purposes. 


This is due to the fact that skin tones, especially darker ones, when in RGB space, tend towards the Red plane. Moving skin tones to HSV makes for a more realistic grouping of skin tones,  as one may infer from below from seeing the Cuties poster color space in different projections. This discussion can be found in this article which looks at the color quantization within the Netflix Cutie's poster and the importance of visual communication with color and other motifs. 


A further example of looking at color in visual communication using a further methodology can be found in this article regarding a Pictorial Analysis of Disney Female Characters.

Research from this is also being used in Lutz's thesis A Counting: Sign Language in which color quantization is utilized to provide automatic white balance to video clips of a variety of resolutions, lighting conditions, and skin tones. 

Current methods often cast a blue or green tone on human subjects, especially ones with darker skin tones.

This is opposed to the default algorithms that are currently available, as seen with the video sample from above when run through a recent  Samsung Research color balancing paper.

Overall, this line of inquiry seeks to understand that we cannot undo the quantification of the human form, but there are more responsible ways to establish how we enact diverse visual representation, and image processing and color quantification are just one of them.

This is an ongoing line of work, please contact Nina Lutz (nlutz@mit.edu) for inquiries.

Project at a glance

Person People
Nina Lutz
Research Assistant