System Information

C++, OpenCV 4.12, Windows, Visual Studio 2022

Detailed description

I upgraded OpenCV from 4.8 to 4.12. After the upgrade, my unit tests for camera calibration started failing because findChessboardCorners(...) now consistently returns false on the exact same input images and parameters.

Steps to reproduce

Mat img = imread("inputimage.jpg");
Size patternSize(13, 9);
vector<Point2f> corners;
int flags = CV_CALIB_CB_ADAPTIVE_THRESH | CV_CALIB_CB_FAST_CHECK | CV_LIB_CB_NORMALIZE_IMAGE;

bool found = findChessboardCorners(img, patternSize, corners, flags);

drawChessboardCorners(img, patternSize, corners, found);
imshow("img", img);
waitKey();

input image

expected result

Code runs fine with 4.8, findChessboardCorners returns false with 4.12

When I resize the input image down to 266×200 pixels before calling findChessboardCorners, I get true again — but the detected corners are much less precise (and not usable for my use case).

Seems to be related to #27885

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
• I updated to the latest OpenCV version and the issue is still there
• There is reproducer code and related data files (videos, images, onnx, etc)