How to define clear range for pixel color

Question

What I'm trying to do is to test if a pixel is blue or not.

For example: The blue color is in RGB defined as rgb(0,0,255). Typical color depths are 8 bit (256 colours), 16 bit (about 65 thousand), 24 bit (about 16 million) and 32 bit (over 4 billion different colours). So there is clearly more than 1 shade of blue.

How do I define the range of the blue color and test for each pixel if it's blue or not? And what do i need to bear in mind regarding the different depths?

My code so far is:

BufferedImage image = ImageIO.read(file);

// Getting pixel color by position x and y
for (int i = 0; i < image.getWidth(); i++) {
    for (int j = 0; j < image.getHeight(); j++) {
        int clr = image.getRGB(i, j);

Note 1:

http://www.workwithcolor.com/cyan-blue-color-hue-range-01.htm

The problem here is, what is in between the color steps?

An example would be great.

Note 2:

I just found a presentation about the topic i'm interested in:

http://cstwiki.wtb.tue.nl/images/06-opencv.pdf

On page 13 we can see the definition of red. But how can you define the other colors?

Answer 1

In my opinion the easiest way is to convert color format to HSV/HSB and then pick blue range for Hue. The exact range may be a problem, but I mostly use from 165 (below is green) to 240 (above is purple/violet). Saturation and Value/Brightness controls only shades, but be aware that Saturation near 0 would be rather gray with some blueish shade, and Value near 0 would be almost black. Because of that you can check also Saturation to be >10 and Value >15, but of course you can pick your own values to suit your needs.

Sample code (not tested):

boolean isBlue(int rgb){
    Color color = new Color(rgb);
    float[] hsb = Color.RGBtoHSB(color.getRed(), color.getGreen(), color.getBlue(), null);
    // now [0] is hue, [1] saturation and [2] value/brightness
    // all values are between 0 and 1 so you must divide hue by 360 and others by 100
    if(hsb[0] > 165/360 && hsb[0] < 240/360 && hsb[1] > 0.1 && hsb[2] > 0.15)
        return true;
    return false;
}

Wiki about HSV

RGBtoHSB javadoc

Color picker to check colors in different systems

Answer 2

In the RGB color space this is not easy to do. For one, the way that colors are displayed is device-dependent, that is, what ever RGB-value you choose to display might look quite different on different screens. But even when this can somehow be controlled for, the colors that have been assigned specific names in English do not necessarily occupy the same volume within the RGB color cube, neither in terms of size nor shape. Because of this, there are a few color models that try to approximate the human perception of colors.

One contender for this is the CIELab color space. The "Lab" refers to how colors are represented in this model: via lightness L of the color, the color's position on the red/magenta and green axis (a), and its position on the yellow and blue axis (b). Unfortunately though, there are no simple formulas for conversion between RGB or CMYK values and Lab*, because the RGB and CMYK color models are device-dependent. The RGB or CMYK values first must be transformed to a specific absolute color space, such as sRGB or Adobe RGB. This adjustment will be device-dependent, but the resulting data from the transform will be device-independent.