Why Does My cv::warpPerspective Deskewing Implementation Produce Incorrect Results, and How Can I Fix It?

How to Correctly Execute cv::warpPerspective for Deskewing a Set of Points

Problem:

Attempts to use cv::warpPerspective to achieve a deskewing effect on a set of points have yielded unsatisfactory results. The desired deskewing is illustrated by the green rectangle in the image below:

[Image of a document with a green rectangle outlining the region of interest]

Cause:

The incorrect results can be attributed to several factors:

1. Point Order: The points in the input and output vectors must be in the same order (e.g., top-left, bottom-left, bottom-right, top-right).
2. Output Image Size: To prevent the resulting image from containing excess background, its width and height should be set to match those of the bounding rectangle around the deskewed region.

Solution:

To resolve these issues, the code below has been modified:

void main()
{
    cv::Mat src = cv::imread("r8fmh.jpg", 1);

    // Points representing the corners of the paper in the picture:
    vector<Point> not_a_rect_shape;
    not_a_rect_shape.push_back(Point(408, 69));
    not_a_rect_shape.push_back(Point(72, 2186));
    not_a_rect_shape.push_back(Point(1584, 2426));
    not_a_rect_shape.push_back(Point(1912, 291));

    // Assemble a rotated rectangle from the points
    RotatedRect box = minAreaRect(cv::Mat(not_a_rect_shape));

    // Extract the corner points of the rotated rectangle
    Point2f pts[4];
    box.points(pts);

    // Define the vertices for the warp transformation
    Point2f src_vertices[3];
    src_vertices[0] = pts[0];
    src_vertices[1] = pts[1];
    src_vertices[2] = pts[3];

    Point2f dst_vertices[3];
    dst_vertices[0] = Point(0, 0);
    dst_vertices[1] = Point(box.boundingRect().width - 1, 0);
    dst_vertices[2] = Point(0, box.boundingRect().height - 1);

    // Use the affine transform as it's faster for the given use case
    Mat warpAffineMatrix = getAffineTransform(src_vertices, dst_vertices);

    cv::Mat rotated;
    cv::Size size(box.boundingRect().width, box.boundingRect().height);
    warpAffine(src, rotated, warpAffineMatrix, size, INTER_LINEAR, BORDER_CONSTANT);

    imwrite("rotated.jpg", rotated);
}

Improvements:

To further enhance efficiency, consider using cv::getAffineTransform() and cv::warpAffine() instead of cv::getPerspectiveTransform() and cv::warpPerspective(). These functions are specifically designed for affine transformations and are significantly faster.

The above is the detailed content of Why Does My cv::warpPerspective Deskewing Implementation Produce Incorrect Results, and How Can I Fix It?. For more information, please follow other related articles on the PHP Chinese website!