After some experimenting, the culprit was the MagnificationFilter. Although there were no problem setting the Magnification with QOpenGLTexture::NearestMipMapNearest with float images, with integer images, I had to set to QOpenGLTexture::Nearest.