Color Matching by Phone

Technology from Hewlett-Packard could help shoppers find the right color.

Hewlett-Packard (HP) Labs has unveiled a research project that could help people select colors when shopping for products. The technology uses color-correcting and computer-vision algorithms that reside on HP servers. The idea, explains Nina Bhatti, principal scientist at HP Labs, is that consumers use their mobile phones to take pictures of themselves or objects, and then send these pictures to HP servers. Within seconds, the consumer receives a text message with a color recommendation for matching makeup to skin tone, or for finding the right paint hue for the home.

Blending in: A woman (above) holds a color-reference chart while her picture is taken with a cell phone. The picture is sent to HP servers, where the chart helps determine the color of the woman’s skin irrespective of lighting and camera quality. Software in the server matches her skin tone to one stored in a database and recommends the appropriate makeup color via text message. Credit: HP

The research prototype that Bhatti and her team developed is specifically designed to help people find the best shade of makeup based on their skin tone. Not all women who buy makeup purchase it from a cosmetics counter, where an expert can identify the best shade of foundation for them, Bhatti explains. Many buy cosmetics from drugstores, catalogues, and, increasingly, the Web, where there's no opportunity to test how the color blends with the skin. "Studies have shown that up to 75 percent of women are wearing the wrong shade of makeup," Bhatti says.

Story continues below

Using HP's prototype, a consumer simply takes a picture of herself while holding a color-reference chart that could be provided at the makeup counter at stores, in catalogues, or in magazines. When the picture is uploaded to HP servers, software compares the values of the color-reference chart in the picture with the accepted values for these colors. The color-correction algorithm takes the difference between these values and applies it to all the pixels in the picture to eliminate the effects of harsh lighting or poor camera quality.

Once the color in the picture is corrected, face-detection software finds the person's face and pulls out its predominant color, overlooking blemishes and other irregularities. This corrected face color is then compared with a database of faces, explains Bhatti. HP researchers have a database of 260 women with different skin tones on whom makeup artists have tested products. The consumer receives a text message identifying the makeup color that looks best on the woman in the database with the most similar skin tone. The technology is described in detail in an upcoming issue of the International Journal of Imaging Systems and Technology. Of course, in the real world, cosmetics companies might want to provide the database of faces, using their own makeup artists to determine the matching hue, says Bhatti.