Converting equirectangular and cubemap image to perspective view image

A perspective image is usually considered with a 90 ° field of view but the code supports other values of FOV in the horizontal and vertical direction. One of the most common applications of this type of conversion is to create a 360 ° image/video viewer, we can also use this method to generate multiple novel pinhole camera images from a single 360 ° image. This is also useful when generating a VR experience or game using skyboxes.

A simplified version of the math behind such conversions is explained in the following figure, to know more mathematical details refer to [1].

Python Example

Example code for equirectangular to perspective conversion

fisheyeUtils.eqruirect2persp(img, FOV, Theta, Phi, Hd, Wd)

Parameters

• img – input cubemap image

• FOV – field of view (in horizontal direction)

• Theta – Viewing angle wrt z axis

• Phi – VIewing angle wrt y axis

• Hd – Output image height

• Wd – Output image width

#!/usr/bin/env/python
import cv2
import numpy as np
import math
import time
import sys
from omnicv import fisheyeImgConv


# path to the input equirectangular image
Img_path = sys.argv[1]

equiRect = cv2.imread(Img_path)
cv2.namedWindow("cubemap",cv2.WINDOW_NORMAL)
cv2.imshow("cubemap",equiRect)
cv2.waitKey(0)

outShape = [400,400]
inShape = equiRect.shape[:2]
mapper = fisheyeImgConv()

FOV = 90
Theta = 0
Phi = 0
Hd = outShape[0]
Wd = outShape[1]
persp = mapper.eqruirect2persp(equiRect,FOV,Theta,Phi,Hd,Wd)
cv2.imshow("cubemap",persp)
cv2.waitKey(0)

Example code for cubemap to perspective conversion

fisheyeImgConv.cubemap2persp(self, img, FOV, Theta, Phi, Hd, Wd)

Parameters

• img – Input equirectangular image

• FOV – Field of view (in horizontal direction)

• Theta – Viewing angle wrt z axis

• Phi – Viewing angle wrt y axis

• Hd – Output image height

• Wd – Output image width

#!/usr/bin/env/python
import cv2
import numpy as np
import math
import time
import sys
from omnicv import fisheyeImgConv

Img_path = sys.argv[1]

equiRect = cv2.imread(Img_path)
cv2.namedWindow("cubemap",cv2.WINDOW_NORMAL)
cv2.imshow("cubemap",equiRect)
cv2.waitKey(0)
outShape = [400,400]
inShape = equiRect.shape[:2]
mapper = fisheyeImgConv()

FOV = 90
Theta = 0
Phi = 0
Hd = outShape[0]
Wd = outShape[1]
persp = mapper.cubemap2persp(equiRect,FOV,Theta,Phi,Hd,Wd)
cv2.imshow("cubemap",persp)
cv2.waitKey(0)

C++ Example

Example code for equirectangular to perspective conversion

fisheyeImgConv::equirect2persp(const cv::Mat &img, cv::Mat &dstFrame,float FOV,

float THETA, float PHI, int Hd, int Wd);

Parameters

• img – input cubemap image

• FOV – field of view (in horizontal direction)

• THETA – Viewing angle wrt z axis

• PHI – VIewing angle wrt y axis

• Hd – Output image height

• Wd – Output image width

• dstFrame – Output images

#include<iostream>
#include<opencv2/opencv.hpp>
#include"../omnicv/utils.hpp"
#include <opencv2/core/core.hpp>

// Creating the display window
int H = 500;
int W = 500;

std::string WINDOW_NAME{"viewer"};

int main()
{
  cv::namedWindow(WINDOW_NAME,CV_WINDOW_NORMAL);

  cv::Mat frame;
  cv::Mat outFrame;

  frame = cv::imread("../data/equirect_temp1.jpg");
  cv::imshow(WINDOW_NAME,frame);
  cv::waitKey(0);
  fisheyeImgConv mapper1;
  mapper1.equirect2persp(frame,outFrame,90,120,45, 400,400);


  return 0;
}

Example code for cubemap to perspective conversion

fisheyeImgConv::cubemap2persp(const cv::Mat &img1, cv::Mat &dstFrame,float FOV,

float THETA, float PHI, int Hd, int Wd);

Parameters

• img – Input equirectangular image

• FOV – Field of view (in horizontal direction)

• Theta – Viewing angle wrt z axis

• Phi – Viewing angle wrt y axis

• Hd – Output image height

• Wd – Output image width

• dstFrame – Output image

#include<iostream>
#include<opencv2/opencv.hpp>
#include"../omnicv/utils.hpp"
#include <opencv2/core/core.hpp>

// Creating the display window
int H = 500;
int W = 500;

std::string WINDOW_NAME{"viewer"};

int main()
{
  cv::namedWindow(WINDOW_NAME,CV_WINDOW_NORMAL);
  // cv::resizeWindow(WINDOW_NAME, 400, 400);

  cv::Mat frame;
  cv::Mat outFrame;

  frame = cv::imread("../data/cubemap_dice.jpg");
  cv::imshow(WINDOW_NAME,frame);
  cv::waitKey(0);
  fisheyeImgConv mapper1;

  mapper1.cubemap2persp(frame,outFrame,90,120,45, 400,400);

  return 0;
}

For a more detailed example refer to the following code. for equirectangular to cubemap projection and refer this code.

Equirectangular to perspective image	Cubemap to perspective image

Reference

[1] Araújo, António. (2018). Drawing Equirectangular VR Panoramas with Ruler, Compass, and Protractor. Journal of Science and Technology of the Arts. 10. 10.7559/citarj.v10i1.471.

Back to examples index