How to find depth map image by using a single image......?????
    9 views (last 30 days)
  
       Show older comments
    
We did till this ... but we are not getting how to produce depth map image ...
close all;
clear all;
clc;
a=imread('cameraman.jpg');
a=imresize(a,[255 255])
a=double(a);
[row col dim]=size(a);
red=a(:,:,1);
[row col]=size(red);
green=a(:,:,2);
blue=a(:,:,3);
numer=0.5*((red-green)+(red-blue));
deno=sqrt((red-green).^2+(red-blue).*(green-blue));
theta=acos(numer./(deno+eps));
%for computing HUE
for x=1:1:row
    for y=1:1:col
        if green(x,y)<blue(x,y);
            H(x,y)=(2*pi)-theta(x,y);
        else
            H(x,y)=theta(x,y);
        end
    end
end
%computing SATURATION & INTENSITY
numer=min(min(red,green),blue);
deno=red+green+blue;
deno(deno==0)=eps;
S=1-(3.*numer./deno);
I=(red+green+blue)/3;
z=edge(I,'canny');
y=imfill(z,'holes');
subplot(2,3,1);
imshow(uint8(a));
colormap(gray);
title('original image');
subplot(2,3,2);
imagesc(H);
colormap(gray);
title('HUE');
subplot(2,3,3);
imagesc(S);
colormap(gray);
title('SATURATION');
subplot(2,3,4);
imagesc(I)
colormap(gray);
title('INTENSITY');
subplot(2,3,5);
imshow(z);
title('edge detection');
subplot(2,3,6);
imshow(y);
title('HOLE FILLING');
figure;
imshow(z);
figure;
imshow(y);
0 Comments
Answers (3)
  Image Analyst
      
      
 on 8 Apr 2017
        
      Edited: Image Analyst
      
      
 on 8 Apr 2017
  
      I don't think you can. Why do you think you can get a depth (range) image from a single color optical photo?
Attach your color cameraman.jpg photo. The standard one that ships with MATLAB is a tiff image and it's monochrome/grayscale, not color. I see no reason why doing edge detection and hole filling would/could get you the distance from the sensor to the subject unless you're somehow hoping that there will be some lateral chromatic aberration.
2 Comments
  Image Analyst
      
      
 on 13 Jan 2020
				Like I said, I don't think it can be done, so there is not code for it.
  George Abrahams
      
 on 23 Jan 2024
        This task is called "monocular depth estimation" and it is a heavily researched field.
For a still image, the most common, traditional method would be shape from shading, but these days deep neural networks are most commonly used. There are many CNNs pretrained for different types of environment freely available.
0 Comments
See Also
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!


