In this paper, we present a method to remove specularities from imaging spectroscopy data. We do this by making use of the dichromatic model so as to cast the problem in a linear regression setting. We do this so as to employ the average radiance for each pixel as a means to map the spectra onto a two-dimensional space. This permits the use of an entropy minimisation approach so as to recover the slope of a line described by a linear regressor. We show how this slope can be used to recover the specular coefficient in the dichromatic model and provide experiments on real-world imaging spectroscopy data. We also provide comparison with an alternative and effect a quantitative analysis that shows our method is robust to changes the degree of specularity of the image or the location of the light source in the scene.