RGB to Grayscale without using rgb2gray

I need to convert RGB image to grayscale without using rgb2gray pre-made code.

I am already getting an error saying "Index in position 3 exceeds array bounds. Index must not exceed 1."

How do I convert RGB to grayscale without rgb2gray? Also, how do I convert back to rgb from grayscale?

RGB_image = ('peppers.png');
r = RGB_image(:,:,1);
g = RGB_image(:,:,2);
b = RGB_image(:,:,3);

Answer 1

[Dave B]

You can choose many approaches for converting to grayscale. For instance you could weight the three colors equally:

im=imread('peppers.png');
im_gray = mean(im,3);                     % take the average of red, green, blue values
im_gray = repmat(uint8(im_gray),[1 1 3]); % convert to integers and fill red, green, and blue values with the grayscale values
imshow(im_gray)

The weighting that im2gray uses is on the documentation page, it over-weights green and under-weights blue.

You can't take a converted grayscale image back to its RGB original, the color information is gone. However there are lots of deep learning algorithms for colorizing. You could search for "colorize deep learning" etc.

Answer 2

[yanqi liu]

convert RGB to grayscale without rgb2gray

convert back to rgb from grayscale

clc; clear all; close all;
RGB_image = imread('peppers.png');
RGB_image = double(RGB_image);
% split rgb channel
R = RGB_image(:,:,1);
G = RGB_image(:,:,2);
B = RGB_image(:,:,3);
% use weight, self define
gray = 0.299 * R + 0.587 * G + 0.114 * B;
% gray to rgb, self define
rgb = zeros(size(gray,1), size(gray,2), 3);
rgb(:,:,1) = gray;
rgb(:,:,2) = 0.5*gray;
rgb(:,:,3) = 0.8*gray;
% display
figure;
subplot(1, 3, 1); imshow(mat2gray(RGB_image)); title('origin rgb');
subplot(1, 3, 2); imshow(mat2gray(gray)); title('gray');
subplot(1, 3, 3); imshow(mat2gray(rgb)); title('gray to rgb');

Answer 3

[DGM]

This answer covers conversions for both BT601 luma (what rgb2gray() and im2gray() use) and BT709 luma.

This answer covers conversions to other grayscale representations of color images (I, V, L, Y, L*)

Alternatively, MIMT mono() can simply return any of the above, without the need for IPT or tedious conversions.