From the view point of sensory biology, the evolution of any modality must correspond with the environment in which it is used. The visual system designs of animals with color vision must take into account the spectral components of the scenes the animals view. A CCD camera coupled with a variable interference filter was used as an imaging spectroradiometer to collect multispectral imagesfrom coral reef scenes. These multispectral images then were transformed into many different receptoral images, based on the modelled rhodopsin absorption spectra (lambda maximum ranged from 400 to 600 nm, with 10 nm interval). Each possible triplet of receptoral images (with three different lambda maximum) were decorrelated by Principal Component Analysis (PCA) to separate color information from intensity information. The expected optimal photoreceptor set is the three receptoral image combination which produces maximal eigenvalues of second and third components of PCA, since the second and third components of PCA represent the spectral variations in the inputs of three photoreceptors.