QImage, parsing raw char from image to get alpha?

[Dariusz]

Hey

I'm doing some openGl rendering and on finish render I would like to get some data out of the image.

The QOpenGLFramebufferObject.toImage() returns a Format_ARGB32_Premultiplied and I use this for format > fboFormat.setAttachment(QOpenGLFramebufferObject::Depth);

Here is the function I'm trying to run but sadly no luck...

    for (int he = 0; he < height; ++he) {
        for (int wi = 0; wi < width; ++wi) {
            pixelOffset = imageWidth * he + (wi * 4); /// get item 0 every 4 items for alpha, since its ARGB, if RGBA mode then ```imageWidth * he + (wi * 4) +4 ```
            alpha = temp[pixelOffset];//qAlpha(img.pixel(wi, he)); /// img.pixel work but slow due to different access function
            im[heightOffset][widthOffset] = alpha;
            widthOffset++;
        }
        widthOffset = 0;
        heightOffset++;
    }

Can any1 help ?
TIA

[mranger90]

What is temp[] ? is it a byte array or a pixel array ?

[Dariusz]

@mranger90 Oh sorry, missed that!

uchar *temp = (uchar *) img.constBits();

[Christian Ehrlicher]

@Dariusz said in QImage, parsing raw char from image to get alpha?:

run but sadly no luck...

What does this mean?

[Dariusz]

That this

uchar *temp = (uchar *) img.constBits();

...
            pixelOffset = imageWidth * he + (wi * 4);
            alpha = temp[pixelOffset];//qAlpha(img.pixel(wi, he)); /// img.pixel work but slow due to different access function

does not return alpha int. as far as I understand img.constBits() return 1d array, 4 values equal to a color, esentially A-R-G-B, so every 4th item is the alpha value and pixelOffset uses/calculates indexOffset in to the 1d array from 2d array of height*width of image.

The loop function in 1st post shows it.

I don't understand why the value I get from the loop above is not alpha.

[Christian Ehrlicher]

Your offset calculation is wrong, every pixel has 4 bytes.

[Dariusz]

@Christian-Ehrlicher said in QImage, parsing raw char from image to get alpha?:

Your offset calculation is wrong, every pixel has 4 bytes.

Hmm did I not address that in (wi*4) ? You say bytes but isn't uchar *temp an array of ints? I'm lost o.o

[Christian Ehrlicher]

@Dariusz said in QImage, parsing raw char from image to get alpha?:

uchar *

I don't see int here... only (unsigned) char

[Dariusz]

@Christian-Ehrlicher Yeah, sorry just used to ints.

So the:
uchar *temp = (uchar *) img.constBits();
Should be something like
temp = [alpha, red, green, blue, alpha, red, green, blue, alpha, red, green, blue, alpha, red, green, blue, alpha, red, green, blue ]
Each value being char. 0-255 - not quite sure how to deal with 32 bit+ arrays tho... are they then int *temp ? constBits return only chars... I'm lost I never dig in to 16/32+ image formats...

so if I were to do temp[0] / temp[4],temp[8] then they all should be alphas?

So what am I missing in
imageWidth * imageHeightIndex + (imageWidthIndex *4)
Which if gives correct numbers would result for 300x400 image (w x h)

300 * 0 + (0*4) = 0
300 * 0 + (1*4) = 4
300 * 0 + (2*4) = 8

etc etc so I'm getting correct index to retrieve alpha ?

What am I missing ? o.O

[Christian Ehrlicher]

@Dariusz said in QImage, parsing raw char from image to get alpha?:

What am I missing ? o.O

Simple math I would say:
4 bytes per pixel, row is 300 pixels = 1200 bytes per row

[Dariusz]

@Christian-Ehrlicher Hmm char is 2 byte big... so do I offset by 8 instead and add up 2 items together?

int alpha = temp[0]<<temp[1] ?

[Christian Ehrlicher]

@Dariusz said in QImage, parsing raw char from image to get alpha?:

char is 2 byte big

Ok, now I'm giving up...

[Dariusz]

@Christian-Ehrlicher said in QImage, parsing raw char from image to get alpha?:

@Dariusz said in QImage, parsing raw char from image to get alpha?:

char is 2 byte big

Ok, now I'm giving up...

Eh crap its short thats 2 byte big... char is 1 byte... ehhhhhhhh I forgot about shorts :/. Ok so esentially I have to grab 4 chars and add them up to int?

[SGaist]

Hi,

Why do you want to combine them back into and int ?

If you have a ARGB image, basically you have:

uchar *start = (uchar *) img.constBits();
uchar *alpha = start + 0;
uchar *red = start + 1;
uchar *green = start + 2;
uchar *blue = start + 3;

// First alpha value:
uint8 alphaValue = *alpha;
// Second value of alpha
alphaValue = *(alpha + 4)

if you want to sum them, then make alphaValue an int and that's all.