Let me break down your problem and share practical solutions based on my experience with camera control on Windows and OpenCV:

The Core Issue #

You've got a Logitech BRIO 4K that works perfectly with OpenCV 3.4.4 + V4L backend on Ubuntu 18.04—you can disable autofocus, lock exposure, and set resolutions without a hitch. But on Windows 10, no matter which OpenCV setup you try (opencv-contrib-python, source builds of 3.4.3/4.1 with MSMF/DSHOW backends), basic capture works fine, but you can't disable autofocus. Even cap.set(cv.CAP_PROP_SETTINGS, 1) fails on both systems, and you've confirmed the correct backend is in use with OPENCV_VIDEOIO_DEBUG=True.

Why This Happens #

As you suspected, the problem lies with OpenCV's Windows video I/O backends:

• MSMF (Media Foundation) and DSHOW (DirectShow) have limited support for advanced camera hardware controls. Logitech's high-end cams like the BRIO expose many parameters that these generic backends don't handle via OpenCV's standard cap.set() interface.
• Linux's V4L2 is a purpose-built framework for video device control, which is why it works seamlessly there—it's designed to talk directly to camera hardware for low-level settings.

Recommended Solutions #

1. Use Logitech's Official Software to Lock Settings (Quick Fix) #

The easiest workaround is to use Logitech's own tools (Logitech Capture or Logitech Camera Settings) to manually disable autofocus, set fixed exposure, and configure other parameters. These settings are saved to the camera's hardware, so when you launch your OpenCV script later, it will inherit the pre-configured values. No code changes needed—perfect if you don't need to adjust settings dynamically.

2. Try the FFmpeg Backend for OpenCV #

If you're willing to recompile OpenCV, enabling the FFmpeg backend can sometimes improve parameter control on Windows. Here's how:

• First, install FFmpeg development libraries on your system.
• When compiling OpenCV from source, make sure to enable the WITH_FFMPEG flag in your CMake configuration.
• In your code, specify the FFmpeg backend when initializing the capture:

  import cv2 as cv
  cap = cv.VideoCapture(0 + cv.CAP_FFMPEG)
  

FFmpeg has better support for a wider range of video devices and parameters than MSMF/DSHOW in many cases—test if it lets you control the BRIO's autofocus.

3. Directly Control the Camera via DirectShow APIs (Advanced) #

OpenCV's standard interfaces don't expose all DirectShow capabilities, but you can use Python's pywin32 library to call DirectShow's low-level APIs directly. Specifically, the IAMCameraControl interface lets you set focus, exposure, and other hardware parameters.

This requires some knowledge of DirectShow, but it's the most reliable way to get granular control on Windows. You can find code snippets that demonstrate how to use pywin32 to interact with camera controls—look for examples using IAMCameraControl::Set.

4. Try Third-Party Enhanced OpenCV Backends #

There are community-modified OpenCV builds that improve Windows camera control. For example, some enhanced MSMF backends add better support for hardware parameters. You can find these modified builds on code hosting platforms or compile them from source to see if they resolve your issue.

Additional Notes #

• The cap.set(cv.CAP_PROP_SETTINGS, 1) function is hit-or-miss on Windows—most backends don't implement the native settings window popup, so it's not surprising it fails for you.
• Logitech's BRIO has a lot of advanced features that rely on proprietary driver hooks, which is why generic frameworks like OpenCV struggle with them on Windows. Using the official software or DirectShow directly will give you the best results.

内容的提问来源于stack exchange，提问作者nick