Color correction describes the transformation process between device response values, e.g., RGB, and CIE XYZ or CIE L*a*b* values, respectively. In every metamer color reproduction system this is the first color transformation after image acquisition. In general, this mapping is not unique because the spectral sensitivities of the majority of devices do not satisfy the Luther condition and the acquisition and viewing light sources typically have different power spectra. Hence, there exists a set of colors with different reflectance spectra which result in the same device response (device metamerism), but have different CIE XYZ tristimulus values for an observer under the viewing light source. In this publication we present a new method to determine this metamer subspace (which depends on the device response) in the viewing CIE L*a*b* space by calculating a Metamer Boundary Descriptor (MBD) matrix. This MBD matrix approximately describes the metamer subspace by storing its boundary points in every entry. By calculating the center of gravity of this MBD we obtain a good color choice in the sense of a small mean error. To calculate the entries of the MBD we use a priori knowledge about the physics of natural reflectance spectra and a linear programming technique. This method improves the performance of target- and regression-based methods especially for saturated colors. We present the results of simulation experiments including a comparison with existing methods.