Question about how to convert a line image to a circle image

Question

First of all, I'm sorry I can not speak English well.

I ask for your understanding even if there are some awkward parts of my words.

I am a beginner to get into Opencv.

I want to translate a picture of a horizontal line into a picture of a circle.

I tried to recommend linearPolar, logPolar, and warpPolar in several articles, but I did not get the desired result.

Result Image:

Here is the image I want to put and the desired result.

Input Image:

Output Image:

What function should I use to get this result?

If you have any examples or explanations for reference, please let me know.

If you let me know just the keywords, I will search Google and this site hard.

Thank you for reading my question.

Answer 1

linearPolar or warpPolar are exactly what you are looking for

You just need to pass it the WARP_INVERSE_MAP flag to make it transform the direction you're looking for.

modified example from samples/cpp/polar_transforms.cpp

Would be something like this:

#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>

int main(int argc, const char** argv)
{
    using namespace cv;
    using namespace std;

    int flags = INTER_LINEAR + WARP_FILL_OUTLIERS;

    Mat src = imread(R"(C:\temp\IPpzR.png)");
    Mat lin_polar_img;
    rotate(src, lin_polar_img, cv::ROTATE_90_COUNTERCLOCKWISE);
    imshow("Loaded",src);
    Mat recovered_lin_polar_img;
    Point2f center( (float)src.cols / 2, (float)src.rows / 2 );
    double maxRadius = min(center.y, center.x);

    cv::linearPolar(lin_polar_img, recovered_lin_polar_img, center, maxRadius, flags + WARP_INVERSE_MAP);
    cv::imshow("Modified",recovered_lin_polar_img);
    cv::waitKey();

    return 0;
}

which gives the following result:

Answer 2

Peter's answer is great (+1).

I had a play with it and just wanted to post a Python version of his approach with some sliders:

#!/usr/bin/env python

import numpy as np
import cv2
# main window
WINDOW_NAME = "carToPol"
cv2.namedWindow(WINDOW_NAME)
# read input
src  = cv2.imread("iPpzR.png")
# get dimenions and compute half (for center/radius)
dims = src.shape[:-1]
cols,rows = dims
half = cols // 2

def carToPol(src,center,maxRadius,interpolation,rotate90=True):
    # rotate 90 degrees (cv2.warpAffine seems to mess up a column)
    if(rotate90):
        src = np.rot90(src)
    # cartesian to polar (WARP_INVERSE_MAP)
    return cv2.linearPolar(src,(centerX,centerY),maxRadius,interpolation + cv2.WARP_FILL_OUTLIERS + cv2.WARP_INVERSE_MAP)

# interpolation names: just for debugging
interpolations = ["NEAREST","LINEAR","CUBIC","AREA","LANCZOS4"]

# slider callbacks: just for debugging
def printCenterX(x):
    print("center",x)
def printCenterY(x):
    print("centerY",x)
def printMaxRadius(x):
    print("maxRadius",x)
def printInterpolation(x):
    global interpolations
    print("interpolation",interpolations[x])
# create sliders
cv2.createTrackbar("centerX"  ,WINDOW_NAME,half,cols,printCenterX)
cv2.createTrackbar("centerY"  ,WINDOW_NAME,half,cols,printCenterY)
cv2.createTrackbar("maxRadius",WINDOW_NAME,half,cols * 4,printMaxRadius)
cv2.createTrackbar("interpolation",WINDOW_NAME,cv2.INTER_CUBIC,cv2.INTER_LANCZOS4,printInterpolation)

# continously process for quick feedback
while 1:
    # exit on ESC key
    k = cv2.waitKey(1) & 0xFF
    if k == 27:
        break

    # read slider values
    centerX   = cv2.getTrackbarPos("centerX",WINDOW_NAME)
    centerY   = cv2.getTrackbarPos("centerY",WINDOW_NAME)
    maxRadius = cv2.getTrackbarPos("maxRadius",WINDOW_NAME)
    interpolation = cv2.getTrackbarPos("interpolation",WINDOW_NAME)

    dst = carToPol(src,(centerX,centerY),maxRadius,interpolation,True)
    # show result
    cv2.imshow(WINDOW_NAME,dst)
    # save image with 's' key
    if k == ord('s'):
        cv2.imwrite("output.png",dst)

# exit
cv2.destroyAllWindows()