In 2006, Americans bought $9.6 billion worth of clothing online [1]. While this figure may seem large, online purchases represented only 5% of all clothing sales. Why are consumers not buying more of their clothes online, when it offers the advantage of convenience? The answer can be found in the fact that customers returned 30% of all the clothes they bought online: when they got their purchase in the mail, it turned out to not be what they had hoped for, in terms of size, texture, color, and other characteristics. It was this idea that motivated us to undertake this project. We hoped that by finding a way to reliably and accurately display the color of fabrics online, customers could be assured that at least the color of the item they're purchasing is what they will receive in the mail.
We began our project by buying samples of fabrics from Jo-Ann Fabric and Craft, a store that also has a website (joann.com) which allows customers to purchase their fabrics online. While perusing the website, we found that there, too, customers had complaints about the colors of their purchases. One woman, for example, warned, "…If I were to order on-line again, I would probably call and confirm the actual description of the color prior to ordering. The color I thought I was ordering was "nowhere" near the color I received. Just be careful." This only served to fortify our position that the colors of fabrics needed to be more accurately displayed online.
The steps from clothing to display
To put our work in context, it is important to understand the basic steps that occur in the process of taking an image of a piece of fabric and displaying it on a monitor [2,3]. These are outlined in the figure below. The camera takes a picture of the fabric under some type of illumination. It records R, G, and B values at each of its pixels, and puts the values through an inverse gamma to make them linear. Then, the camera performs color balancing, where it adjusts the R, G, and B values to "improve" the image it just captured. Finally, the adjusted R, G, and B values are put through a monitor gamma and displayed on an sRGB monitor.
Errors along the way
Though this process occurs every time a digital picture is taken and downloaded to a computer, there are many errors introduced along the way. For amateur photographers, the errors are often of little consequence, but for a concerned consumer, they can be very important. One source of error is the assumptions made about the monitor. Most cameras assume an sRGB monitor will be used to display the images and that its gamma will be set to 2.2. The sRGB color space is a standard color space adopted by many companies in the 1990s in an attempt to create a default color space for the Internet, many operating systems and other computer-related projects [4]. Though the sRGB space was originally intended for use with cathode ray tube monitors, most of today's monitors can be set to appear like an sRGB monitor. The gamma can usually also be changed to a value of 2.2 However, often the user does not know or care enough to make sure their monitor is calibrated correctly, and thus the images are displayed inaccurately.
Another source of error is one that is common with digital photography, and that is the loss of information due to the limited number of channels on the camera. Digital cameras usually only capture information about three channels (red, green, and blue), and very often information is lost.
A source of error that we looked into specifically in our project was the illuminant assumed or estimated by the camera when the image was captured. In the commonly-used "Auto" setting, digital cameras often estimate what illuminant was used. The cameras can also often be manually set for certain illuminants. The importance of accurately setting this feature is shown below. Pictures of a piece of blue fleece were taken under tungsten light. For each of the three images shown, the setting on the camera was changed to be set for tungsten, fluorescent, or daylight illuminants. As seen here, the setting can greatly change how the image appears.
The importance of the illuminant setting leads to the greater issue of the color balancing done by the camera. The color balancing algorithm followed by digital cameras is often proprietary and not made known to the public. Thus, understanding any processing that it does following image capture is made difficult.
