在 Python 中扫描透视图像

Question

我正在尝试准备几张从不同角度拍摄的乐谱图像，以便我可以进行 OMR。

我尝试过变换和旋转它们，但没有帮助，我想我需要调整图像的视角

我试过了：

def rotate_image(mat, angle):
    height, width = mat.shape[:2] # image shape has 3 dimensions
    image_center = (width/2, height/2) # getRotationMatrix2D needs coordinates in reverse order (width, height) compared to shape

    rotation_mat = cv2.getRotationMatrix2D(image_center, angle, 1.)

    # rotation calculates the cos and sin, taking absolutes of those.
    abs_cos = abs(rotation_mat[0,0])
    abs_sin = abs(rotation_mat[0,1])

    # find the new width and height bounds
    bound_w = int(height * abs_sin + width * abs_cos)
    bound_h = int(height * abs_cos + width * abs_sin)

    # subtract old image center (bringing image back to origo) and adding the new image center coordinates
    rotation_mat[0, 2] += bound_w/2 - image_center[0]
    rotation_mat[1, 2] += bound_h/2 - image_center[1]

    # rotate image with the new bounds and translated rotation matrix
    rotated_mat = cv2.warpAffine(mat, rotation_mat, (bound_w, bound_h), borderValue=[255,255,255,0])
    return rotated_mat

# the first call
def our_rotate(RGBImage):
    angle = getAngle(RGBImage)
    rot = rotate_image(RGBImage,angle)
    return rot

Answer 1

单应性的思想是，给定一组对应点，我们可以计算将第一组点扭曲到第二组所需的矩阵。如果我们挑出乐谱的角并告诉它们变形为一个矩形，那么我们得到的单应矩阵将完全执行该操作。通过将该扭曲应用于所有点，我们可以纠正图像。

结果的好坏取决于页面开始时的平整度、镜头畸变程度（您可以使用 opencv 通过一些棋盘校准来消除畸变）以及选择角落的准确度。

代码如下：

import cv2
import math
import numpy as np

# corner clicks
corners = []; # assuming clockwise from top-left: [top-left, top-right, bottom-right, bottom-left]
def clicky(event, x, y, flags, param):
    # get globals
    global corners;

    # check for mouse click
    if event == cv2.EVENT_LBUTTONUP:
        if len(corners) < 4:
            corners.append((x,y));

# returns 2d distance
def dist2D(one, two):
    dx = one[0] - two[0];
    dy = one[1] - two[1];
    return math.sqrt(dx*dx + dy*dy);


# load image
img = cv2.imread("sheet.png");

# make a set window
cv2.namedWindow("Image")
cv2.setMouseCallback("Image", clicky)

# loop
done = False;
while not done or len(corners) < 4:
    # make a copy
    copy = img.copy();

    # draw corners
    for corner in corners:
        copy = cv2.circle(copy, corner, 4, (0,0,255), -1);

    # show
    cv2.imshow("Image", copy);
    key = cv2.waitKey(1);

    # check keys
    done = key == ord('q');

# destination points (a rectangle)
margin = 100;
half = int(margin / 2);
height = int(dist2D(corners[0], corners[3]));
width = int(dist2D(corners[0], corners[1]));
rect = [];
rect.append([half,half]);
rect.append([half+width,half]);
rect.append([half+width,half+height]);
rect.append([half,half+height]);

# get the homography matrix
corners = np.array(corners);
rect = np.array(rect);
hmat, ret = cv2.findHomography(corners, rect);

# warp image
warped = cv2.warpPerspective(img, hmat, (width+margin, height+margin));

# show
cv2.imshow("Warped", warped);
cv2.imwrite("marked.png", copy);
cv2.imwrite("warped.png", warped);
cv2.waitKey(0);