The basic question is, what factors, physical, biochemical and physiological, set the limits of visual capabilities in the intensity (sensitivity), spatial (acuity), and spectral (wavelength) domains, and how are they related to each other? There is a trade-off between sensitivity, visual acuity and color vision that roots in the quantum nature of light. A certain minimum number of quanta is required to paint a picture with a given sharpness and smooth shades of color. Thus low-light vision can only be improved by collecting more photons, that is by making eyes bigger, increasing the fraction of absorbed light, and adjusting spectral sensitivity of photoreceptors to the spectrum of ambient illumination. However, quantum-limited performance can further be spoiled by the noise originating in the visual system. One of the noise sources is the thermal activation of visual pigment molecules. It results in spontaneous excitation events that occur at random intervals and are identical to those signaling real photon absorptions. They constitute an irreducible noise that is added to the quantum noise of the image itself. Both experimental and theoretical studies show that the visual pigment noise steeply increases with shifting the spectral sensitivity curves towards longer wavelengths. This explains the occurrence of a Purkinje shift during rod- cone transition, and sets the "red" limit of the visible spectrum. Maximizing signal-to-noise ratio can be an important factor that governs the selection of an optimum visual pigment for a given photic environment.