I have written a computer program that can detect coins in a static image (.jpeg, .png, etc.) using some standard techniques for computer vision (Gaussian Blur, thresholding, Hough-Transform etc.). Using the ratios of the coins picked up from a given image, I can establish with good certainty which coins are which. However, I wish to add to my confidence levels and also determine if a coin that I deduce to be of type-A (from radius ratios) is also of the correct colo[u]r. The problem is that for British coins et al. (copper, silver, gold), the respective colors (esp. of copper to gold) are very similar.

I have a routine that extracts the mean color of a given coin in terms of the RedGreenBlue (RGB) 'color-space' and routines to convert this color into HueSaturationBrightness (HSB or HSV) 'color-space'.

RGB is not very nice to work with in attempting to differentiating between the three coin colors (see attached [basic] image for an example). I have the following ranges and typical values for the colours of the different coin types:

Note: the typical value here is one selected using a 'pixel-wise' mean of a real image.

**Copper RGB/HSB:** typicalRGB = (153, 117, 89)/(26, 0.42, 0.60).

**Silver RGB/HSB:** typicalRGB = (174, 176, 180)/(220, 0.03, 0.71).

**Gold RGB/HSB:** typicalRGB = (220, 205, 160)/(45, 0.27, 0.86) 

I first tried to use the 'Euclidian distance' between a given mean coin color (using RGB) and the typical values for each coin type given above treating the RGB values as a vector; for copper we would have:

$$D_{copper} = \sqrt((R_{type} - R_{copper})^{2} + (G_{type} - G_{copper})^{2} + (B_{type} - B_{copper})^{2})$$

where the smallest value of the difference ($D$) would tell us which type the given coin is most likely to be. This method has shown itself to be very inaccurate.

I have also tried just comparing the hue of the coins with the typical values of the types provided above. Although theoretically this provides a much better 'color-space' to deal with varying brightness and saturation levels of the images, it too was not accurate enough.

Question: What is the best method to determine a coins type based on color (from a static image)?

Thanks very much for your time.

Edit

Note: I have tried all of the ideas discussed below and have achieved next to nothing. Variance in lighting conditions (even within the same image) make this problem very tough and should be taken into consideration.