恢复OpenCV中横向全屏的原有比例适配显示行为

我使用Ubuntu 22.04 LTS预装的OpenCV版本，此前通过以下配置设置窗口：

cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)
cv2.setWindowProperty(window_name, cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)

该配置可让宽于屏幕的图片按比例缩放，完整适配横向全屏显示。但安装（及重装）PyQt5、PyQt5-sip等包后，窗口变为纵向拉伸图片导致失真。

以下是最小可复现示例（MRE）：

import cv2
import numpy as np
from screeninfo import get_monitors

def get_screen_size():
    monitor = get_monitors()[0]
    return (monitor.width, monitor.height)

def write_text(image, text="Hello World"):
    font = cv2.FONT_HERSHEY_SIMPLEX
    font_scale = 14
    font_thickness = 14
    font_color = (255, 255, 255)
    (text_width, text_height), _ = cv2.getTextSize(text, font, font_scale, font_thickness)
    x = (image.shape[1] - text_width) // 2
    y = (image.shape[0] + text_height) // 2
    cv2.putText(image, text, (x, y), font, font_scale, font_color, font_thickness)
    return image

image = np.random.randint(0, 256, (1280, 2560, 3), dtype=np.uint8)  # dummy image
image = write_text(image)
window_name = 'Full Screen Image'
cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)
# cv2.resizeWindow(window_name, get_screen_size()) # 1920 x 1200 # This is my workaround. But originally this line was not in the code
cv2.setWindowProperty(window_name, cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
cv2.imshow(window_name, image)
cv2.waitKey(0)
cv2.destroyAllWindows()

如何恢复原有全屏行为，让图片按比例适配屏幕横向显示且不失真？

安装PyQt5后，OpenCV的窗口后端被切换为Qt，导致原有的全屏缩放逻辑失效。可通过以下两种方式恢复按比例缩放的全屏显示：

方法1：强制使用GTK窗口后端 #

在代码开头添加环境变量设置，让OpenCV回到Ubuntu预装版本默认的GTK后端，恢复自动按比例缩放的行为：

import os
# 强制OpenCV使用GTK后端
os.environ['OPENCV_UI_BACKEND'] = 'GTK'

添加后保持原有的窗口配置代码即可，无需额外调整。

方法2：手动计算适配比例（不切换后端） #

如果需要保留Qt后端，可手动计算图片与屏幕的适配比例，先调整窗口尺寸再设置全屏：

import cv2
import numpy as np
from screeninfo import get_monitors

def get_screen_size():
    monitor = get_monitors()[0]
    return (monitor.width, monitor.height)

def write_text(image, text="Hello World"):
    font = cv2.FONT_HERSHEY_SIMPLEX
    font_scale = 14
    font_thickness = 14
    font_color = (255, 255, 255)
    (text_width, text_height), _ = cv2.getTextSize(text, font, font_scale, font_thickness)
    x = (image.shape[1] - text_width) // 2
    y = (image.shape[0] + text_height) // 2
    cv2.putText(image, text, (x, y), font, font_scale, font_color, font_thickness)
    return image

image = np.random.randint(0, 256, (1280, 2560, 3), dtype=np.uint8)
image = write_text(image)
window_name = 'Full Screen Image'

# 获取屏幕与图片尺寸
screen_w, screen_h = get_screen_size()
img_h, img_w = image.shape[:2]

# 计算适配比例，取宽高比例的最小值，确保图片完整显示
scale = min(screen_w / img_w, screen_h / img_h)
# 计算缩放后的窗口尺寸（保持图片比例）
target_w = int(img_w * scale)
target_h = int(img_h * scale)

cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)
cv2.resizeWindow(window_name, target_w, target_h)
cv2.setWindowProperty(window_name, cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
cv2.imshow(window_name, image)
cv2.waitKey(0)
cv2.destroyAllWindows()

原因说明 #

Ubuntu预装的OpenCV默认使用GTK窗口后端，该后端在设置全屏时会自动按图片比例缩放以适配屏幕；而安装PyQt5后，OpenCV会优先切换到Qt后端，Qt后端的全屏逻辑是直接拉伸窗口填充屏幕，不会保持图片比例，因此导致失真。

内容的提问来源于stack exchange，提问作者ravi