How do I get a value from a JObject?

We know that a color image has 3 channels R,G and B. If we are given a pixel value of each R,G and B value, how do we get the color value?

• For eg: in YCbCr color space, Y = 0.299R + 0.587G + 0.114B Similarly, if I want a value for color value at a pixel in the color image using colorValue = x*R + y*G + z*B What will be x,y and z values?

• Answer:

  Whenever talking about color, always remember the first and the most basic lessons of all : Color is a PERCEPTION. Thus, blue is just a perception of some specific set of pixel values coming together. Hence, when it comes to measuring color, your ideal objective is to quantify the perception. Now, that is the part where it gets super-difficult. Many psychophysical experiments have been carried out and they do tell us that color is a too complex phemomenon about which we still have very limited knowledge. Now about the YCbCr color spaces, you should know that the conversion which you have highlighted in your question adheres to a number of assumptions : In short and simple words, it assumes that the RGB primaries used to encode the image follow the CCIR 601 television standard ( http://en.wikipedia.org/wiki/Rec._601 ) . A different television standard like ITU-R BT.709 which is meant for HDTV's would call for a different weighted combination of the R , G and B channels.  Infact, if your display is a digital display the combination would change dramatically because then you would have to care about the number of bits used to encode information by the display. Thus the kind of weighted combinations you are referring to are well thought of and follow certain restrictions.  For different television standards , the same pixel values can show very different colors.(Actually that is one of the ways in which different companies would try to convince you that their display is better than their competitiors !! ) Now if you want to measure color, it is important to understand, that what are you actually looking to measure ? Biologically, there is nothing like a global color measure. Our perception of colors is fuzzy and relative. That is why it is not possible to define the absolute range of a color like "Blue" or "Green". If you look at YCbCr, then Y component is luma component which is not color information. Cb and Cr components measure biologically plausible opponent color quantities. These quantities measure something concrete but even that is remotely related to color since the final perception of color depends on the relative values of Y Cb and Cr amongst other conditions like lighting conditions. Hence, quantifying a single value which represents color as we know it and that through directly through a linear transformation of the pixel values, is in my opinion a not so well-thought of idea. But, yes, in different applications people try to look for certain linear combinations for achieving certain objectives. For example, some people try to look for weighted combinations which would enhance the contrast of an image in a specific way. However there is no single holistic representation of color into a scalar number. Moreover, if you want a more rigorous explanation, then a basic theorem of topology says that you cannot reshape a 3D space like RGB cube to a one-dimensional representation so that the relative spatial configuration of points can be preserved. Hence, you cannot represent color by a single scalar number such that it will encompass the relative perceptual nature of color.