Currently, existing color-capturing systems collect improper information about colors. It is happens because present CCD and CMOS matrixes (sensors plus color filters) have spectrum characteristics that do not match the spectrum characteristics of the human eye. This creates initial error in a color-capturing process: sensors do not capture exactly the same information as eyes do. As a result, there will always be a situation when two colors look alike for the sensors but different for a human eye and vice versa. It is theoretically impossible to distinguish these two colors, as the information, not captured either by a sensor or by an eye, is lost forever.

For example, one of Sony’s CCD sensors has spectrum characteristics similar to shown on the left, while human cone sensitivity is as shown on the right.

An example of two colors that are different to the human eye, but are identical to the sensor of Sony is provided below:

Light consisting of two monochromatic components: 0.628 W with the wavelength λ=400nm and 0.848 W with the wavelength λ=600nm, is visible to a human as a bright orange color. Tristimulus values are (X = 905, Y = 533, Z = 43).

Light consisting of two monochromatic components: 0.233 W with the wavelength λ=500nm and 1.000 W with the wavelength λ=700nm, is visible to a human as a dark green color. Tristimulus values are (X = 12, Y = 79, Z = 63).

However, the sensor of Sony perceives these two lights as identical.
Sensor values for both lights are (B = 138, G = 281, R = 829) =

There is no algorithm that would be able to distinguish the first light from the second, from the sensor’s values. Algorithms can calculate only one color for both cases (not necessarily red or green, it might be yellow, depending on internal algorithm).