Images

[NotAGolf]

I'm doing a java project to do with stereo vision. I'm having problems with regards to getting a colour value from a particular pixel in an image. I need to compare particular pixels between 2 images so to do this i have stored the pixels value in a 2D array using the line of code:

pixels[j] = bufferedImage.getRGB(i,j); (pixels is an integer 2d array)

But this stores a strange value which i dont understand: -5394273
This pixel's value in photoshop is:
R-82
G-60
B-73
# - 523c49 (it's a very dark purple).


What am I doing wrong? Is there another function i could use insetad of getRGB()? I can read the image in as a grayscale image in if that would make things easier.

Thanks for any advice.

 

[DaveF]

Why not use getRed, getGreen, getBlue?

 

[NotAGolf]

I've thought about using getRed etc, but I cant seem to get i to work. This is my method at the moment, the commented out line is how i thought I could use it but it gives the error 'cannot find symbol getRed'.

//Create a 2Darray of pixels of the image
    public void getPixels() {
        
        int width = bufferedImage.getWidth();
        int height = bufferedImage.getHeight();
        pixels = new int[width][height];
        
        for(int i=0; i<width; i++) {
            for(int j=0; j<height; j++) {
                pixelsLeft[i][j] = bufferedImage.getRGB(i,j);
                //pixelsLeft[i][j] = bufferedImage.getRed(i,j);
            }
        }
    }

I'm including the libraries:
import java.awt.*;
import javax.imageio.*;
import java.io.*;
import java.lang.*;
So i think i've got the right libraries and such

Sorry for my noobiness but my lecturer is foreign and I cant understand what he says so I cant go to him for help

 

[DaveF]

I've thought about using getRed etc, but I cant seem to get i to work. This is my method at the moment, the commented out line is how i thought I could use it but it gives the error 'cannot find symbol getRed'.

You could try just shifting and masking to get the RGB components. Something like:

int v = bufferedImage.getRGB(i,j);
int b = v & 255;
int g = (v>>8)&255;
int r = (v>>16)&255;

Dunno if it will work as you want though.

 

[NotAGolf]

You could try just shifting and masking to get the RGB components. Something like:

int v = bufferedImage.getRGB(i,j);
int b = v & 255;
int g = (v>>8)&255;
int r = (v>>16)&255;

Dunno if it will work as you want though.

Am I right in thinking that if you have a grayscale image, the red, green and blue are the same? And i'm really sorry for all the questions, but what does v&255 do? I would search on the web myself but my internet is unbelievably slow and the only only thing that seems to work is OCUK. Nice work guys!!!

 

[DaveF]

Am I right in thinking that if you have a grayscale image, the red, green and blue are the same?

Unless you're doing something really wierd with the colour space, yes.

'&' = logical and. So v&255 is the least significant byte of v.

 

[NotAGolf]

Thanks very much for your help Dave, it seems to bring up more sensible values now. However these values are still quite a way off from the colour values of the pixel in photoshop, is this normal? Surely they should be the same, if not very close?

 

[DaveF]

With images, you'll see people talk about colour spaces. Loosely put, this is all about the concept that R,G,B is "just a set of numbers" that don't necessarily translate into colour intensities in the way you might think.

For example, with a typical CRT monitor (with no gamma correction), you will find that R=0.5 results in a pixel that emits much less than half the light of a pixel with R=1. Some pieces of software use quite extensive algorithms to try to deal with this.

I don't know enough about Photoshop colour spaces to know if that's the issue here (or indeed whether Java may do something similar), but it's a possibility. Or it may be there's something else going on. As long as the values seem to make sense, I wouldn't worry too much.

One check that might reassure you is to try setting the RGB values (in Java), and then reading them back using the code I suggested. If the values match, no problem. If they don't, then more investigation might be needed.

 

[NotAGolf]

Dave=

Thanks for all your help!