How to support high-res pixmap in qgraphicsproxywidget when zooming - qt

Question

I have a qgraphicsview with a scene that contains a qgraphicsproxywidget. The widget currently shows some hi-res images in a square about 25x25 via QPixmap. I'm looking for the proper approach to support zooming in on the image without deteriorating its resolution so much. I've found hints that it might be possible by overriding the paint method (ie derive from QPixmap then override paint method), or by using QImage, or some configuration options (I have tried setSmoothRendering on the QGraphicsView but this only helps a little), but it's not clear if these techniques apply when the image is in a widget in a graphics view.
I wrote the following program (actually, my colleague Colin did, I simplified it for posting) that shows the technique I use. Once you save the attached image and run the program, position the mouse over the pixmap, and press + several times to zoom in: notice how the pixmap is pixelated, whereas the text is perfect.
from PyQt5.QtCore import Qt
from PyQt5.Qt import QPixmap
from PyQt5.QtGui import QPainter
from PyQt5.QtWidgets import (QApplication, QGraphicsScene, QWidget,
QGraphicsView, QGraphicsProxyWidget, QLabel, QVBoxLayout, QHBoxLayout)
class TestGraphicsWidget(QWidget):
def __init__(self):
super().__init__()
self.setupUi(self)
def setupUi(self, TEST):
TEST.resize(400, 300)
self.verticalLayout = QVBoxLayout(TEST)
self.widget = QWidget(TEST)
self.horizontalLayout = QHBoxLayout(self.widget)
self.image_label = QLabel(self.widget)
self.image_label.setStyleSheet("border: 2px solid red;")
self.horizontalLayout.addWidget(self.image_label)
self.text_label1 = QLabel(self.widget)
self.text_label1.setStyleSheet("border: 2px solid red;")
self.horizontalLayout.addWidget(self.text_label1)
self.verticalLayout.addWidget(self.widget)
self.text_label2 = QLabel(TEST)
self.text_label2.setStyleSheet("border: 2px solid red;")
self.verticalLayout.addWidget(self.text_label2)
TEST.setWindowTitle("Form")
self.text_label1.setText("TEXT LABEL 1")
self.text_label2.setText("TEXT LABEL 2")
class TestView(QGraphicsView):
def __init__(self):
super().__init__()
scene = QGraphicsScene(self)
scene.setItemIndexMethod(QGraphicsScene.NoIndex)
scene.setSceneRect(-400, -400, 800, 800)
self.setScene(scene)
self.setCacheMode(QGraphicsView.CacheBackground)
self.setViewportUpdateMode(QGraphicsView.BoundingRectViewportUpdate)
self.setRenderHint(QPainter.Antialiasing)
self.setTransformationAnchor(QGraphicsView.AnchorUnderMouse)
self.setResizeAnchor(QGraphicsView.AnchorViewCenter)
test_widget = TestGraphicsWidget()
test_widget.image_label.setFixedSize(25, 25)
test_widget.image_label.setScaledContents(True)
test_widget.image_label.setPixmap(QPixmap(r"chicken.jpg"))
proxy = QGraphicsProxyWidget()
proxy.setWidget(test_widget)
proxy.setPos(-100, -100)
scene.addItem(proxy)
self.scale(0.8, 0.8)
self.setMinimumSize(400, 400)
def keyPressEvent(self, event):
key = event.key()
if key == Qt.Key_Plus:
self.scaleView(1.2)
elif key == Qt.Key_Minus:
self.scaleView(1 / 1.2)
else:
super().keyPressEvent(event)
def scaleView(self, scaleFactor):
self.scale(scaleFactor, scaleFactor)
if __name__ == '__main__':
import sys
app = QApplication([])
widget = TestView()
widget.show()
sys.exit(app.exec_())
Although this is not all that surprising because the pixmap is downsampled into a 25x25 square of pixels then scaled up by the view, I wouldn't be surprised if there is a qt-specific technique I'm missing, like perhaps something that can be done by overriding the paint of QPixmap to take into account the current scale of the view. I have no choice about using QGraphicsView or embedding an image in a QGraphicsProxyWidget, but I have freedom on image format, the configuration of view or the class to use to load the image into qt, etc.
Any help would be really appreciated.

Answer 1

You appear to be asking: how do I downscale a 600x600 jpg to 25x25 without pixelation? The answer to which is obviously: you can't.
If you put a crappy little 25x25 jpg image in a graphics-view and wind the scale in and out, it's just like standing closer or further away from it. The image doesn't change at all: only your view of it does. And the closer you are to it, the more its intrinsic crappiness is revealed.
So one solution would appear to be: start with a much bigger subject. Resize the widget to, say, four times its original size (and likewise the image label), and then scale down the graphics view to get back to the widget's original starting size.