Color assessment for Home Décor


Joanna Marguier


Digital cameras have never been so common, compact, and affordable. Especially since they have been integrated into cell phones, most of us walk around with a camera at hand at all times. We propose to use cell phone cameras to give color advice while shopping. The user sends an image of an object imaged with an adapted color calibration target and receives information on colors complementing the object.

The RGB values of a pixel depend basically on three elements: the reflectance properties of the imaged object, the lighting conditions, and the camera sensor response. The color properties of an object are fully characterized by its reflectance spectra, i.e. the percentage of light reflected at each wavelength and angle. Colorimetric information cannot be retrieved from an image taken with an uncalibrated camera under unknown illuminants without some additional information.

We present a method performing a post-calibration to colors in images taken with a variety of consumer cameras under uncontrolled illuminant. The object of interest is imaged with a target, whose values are extracted and compared to reference values, allowing to compute an image dependent color correction transform.

We have shown that we can correct a small range of colors with high accuracy, such as skin tones. We extended the method to virtually any color by designing a set of color correction targets covering all hues. For each object, we select a target whose gamut contains the object color. The method was applied to Home Dйcor and the results were validated as very good by a Home Dйcor expert.

Figure 1: An object is imaged with an adapted calibration target. The target patches are extracted and used to compute a color transform applied to the entire image. The second row shows the target gamuts before and after color correction (cyan and magenta surfaces, respectively) along with the reference values (green surface). As long as the object color falls into the target gamut, the color correction is accurate.


HP Laboratories , Palo Alto, CA, USA
Sabine Süsstrunk


HP Laboratories , Palo Alto, CA, USA