Shulou(Shulou.com)06/01 Report--

Python+OpenCV how to achieve automatic poster scene replacement, I believe that many inexperienced people are helpless about this, this article summarizes the causes of the problem and solutions, through this article I hope you can solve this problem.

Now there is a problem, in the case of the two books in the picture below, how to quickly align the book in the middle with the book on the left (the final effect can be achieved by overlapping the two pictures (the final result is right)), and the image transformation can be rotated, translated, scaled, and deformed.

The main content is to introduce how to use Opencv to solve the above problem. Solving this problem requires three steps.

Determine at least four sets of corresponding point coordinates

Find a transformation matrix;

Apply the transformation matrix found to Moving Image to achieve image alignment

The main purpose of image rotation, translation, scaling and other operations is to finally realize the point-to-point mapping relationship between two images. Image mapping is essentially pixel conversion.

Four of the corresponding points are marked in different colors, red, orange, yellow and green; for example, here

and

Yes is a set of corresponding points. After image transformation, points must be mapped to point positions.

The coordinate transformation of points in pictures requires matrix operation. The image dimensions explored here are all two-dimensional, and the coordinates only need to be able to be used.

For this type of transformation problem, the Homography transformation ( 3 × 3 matrix) can be used to solve this type of transformation problem. To solve the point-to-point mapping problem, the Homography matrix can be written as follows:

then

、

As a corresponding point, the application of Homography is as follows:

At least four sets of corresponding points are needed to determine the parameters of matrix H, so at least four sets of corresponding points must be found when calculating H; the more sets of corresponding points are found, the more accurate H will be calculated, and the final conversion effect will be better.

Here is how to align the books in the picture above using Opencv + Python

import cv2

import numpy as np

if __name__ =='__main__':

#Picture reading

img_src = cv2.imread("D:/book2.jpg")

position_src = np.array([[141,131],[480,159],[493,630],[64,601]],dtype = float)

img_dst = cv2.imread("D:/book1.jpg")

position_dst = np.array([[318,256],[543,372],[316,670],[73,473]],dtype = float)

#Calculate the transformation matrix

h,status = cv2.findHomography(position_src,position_dst)

#Affine transformation of pictures

out_img = cv2.warpPerspective(img_src,h,(img_dst.shape[1],img_dst.shape[0]))

#Display images;

cv2.imshow("Source image",img_src)

cv2.imshow("Destination Image",img_dst)

cv2.imshow("Warped Source Image",out_img)

cv2.waitKey(0)

Here, the coordinates of the corresponding four points have been determined in advance, and then the coordinates of these four points are substituted into cv2.findHomography() to calculate the transformation matrix. Finally, the matrix is applied to the two images to obtain the final transformation result.

A reminder here is that when warpPerspective performs matrix transformation on image pixels, it hides a parameter Interpolator, which defaults to linear interpolation and prevents pixel values from missing.

The above small case inconvenient point needs to determine the coordinates of the corresponding four points, this step is more cumbersome, the following case will be added to the program interactive function, to achieve a certain picture automatic marker point collection, marker point conversion:

First of all, you need to prepare two pictures, one of which is a poster and the other is a poster that needs to be replaced. When determining the coordinates of the point, the coordinates of the replaced picture are very easy to determine, just know the length and width of the picture.

However, the four points in the poster image area are difficult to determine. Here, the mouse callback function of Opencv is used to monitor the mouse response and collect the coordinates obtained by PIck according to the user click.

def mouse_handler(event,x,y,flags,data):

if event ==cv2.EVENT_LBUTTONDOWN:

cv2.circle(data['im'],(x,y),3,(0,0,255),5,16)

cv2.namedWindow("Image",0)

cv2.imshow("Image",data['im'])

if len(data['points'])

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.