Normally the pixell array is on RGB form

I would suggest you read the header of the file first , it gives the format of image data, say RGB , YCbCr , Lab so on .  It also has the bpp (bits per pixel representation) say 16 bit of RGB may be (5 - Red , 6 -Green and 5-blue) or RGB(8 bit of each )  . All files have a format of representation so you need to read the metadata (worst case see properties of the file )

Depending on how you want to process , you can make it unsigned integer of 8 bits or combine as one pixel by say x2 * Red + x * Green + Blue

For color space : https://en.wikipedia.org/wiki/CIE_1931_color_space

Hi 

If are working in matlab this link is useful

http://in.mathworks.com/help/images/ref/impixel.html

http://in.mathworks.com/company/newsletters/articles/how-matlab-represents-pixel-colors.html

and in general for image processing you check this

https://processing.org/tutorials/pixels/

You could convert your image to the Lab-Colorspace (or similar as L*a*b*) that is constructed so that the euclidean distance between colors corresponds to the perceived difference of the colors. If the distance is below a threshold, they are seen as the same color and could be counted as one color in your case. A clustering of your pixels in groups where the pairwise distance is below this threshold could give you a list of the different colors and their number. This clustering however is most probably not unique but the question is also not exactly defined.

https://en.wikipedia.org/wiki/Color_difference

write a c++  code demonstrate a hierarchical clustering and import it  to matlab  

same concept of histogram, check if gray level has value or not