Mirror image to align QImage and OpenGL texture coordinates

[alxvth]

In ImageProp::setColorMapImage we use

texture.reset(new QOpenGLTexture(colorMapImage));

Qt's docs on QOpenGLTexture suggest:

QOpenGLTexture *texture = new QOpenGLTexture(QImage(fileName).flipped());

Note that the QImage is flipped vertically to account for the fact that OpenGL and QImage use opposite directions for the y axis. Another option would be to transform your texture coordinates.

Currently this yields a different re-coloring in the ImageViewer vs the Scatterplot for the same data:

The data here is a 2D image where each pixel value is its coordinates, i.e. at (x: 0, y: 0) the value is (1: 0, 2: 0) and at (x: 25, y: 50) the value is (1: 25, 2: 50)

When mirroring the texture image we get consistent behavior:

In the scatterplot we use Texture2D::loadFromImage from the core (Texture.h) via the PointRenderer.
There we just use (QRgb*)image.bits() to point to the color values instead of converting a QImage to QOpenGLTexture, which appears to behave differently.

[alxvth]

Just for sake of reproducibility, the code to generate the test images:

import numpy as np
from PIL import Image

size = 100

# Image 1: x-gradient (columns 0-99, same across all rows)
# Flip vertically so (0,0) is bottom-left
x_array = np.tile(np.arange(size, dtype=np.float32), (size, 1))
x_array = np.flipud(x_array)

# Image 2: y-gradient (rows 0-99, constant across each row)
# y goes 0 (bottom) to 99 (top)
y_values = np.arange(size, dtype=np.float32)
y_array = np.tile(y_values, (size, 1)).T
y_array = np.flipud(y_array)

Image.fromarray(x_array, mode='F').save('./1.tiff')
Image.fromarray(y_array, mode='F').save('./0.tiff')