How to compress 32 bit bmp images to 8 bit bmp?

I have thousands of 32 bit grayscale bmp images that I want to compress to 8 bit bmp grayscale images using a script to reduce the file size and save space while maintaining the image quality. I have looked at using the "im2uint8()", "rgb2ind()", and "imwrite()" functions to try and achieve this, but I either get fully black images or, when it compresses images that maintain the image quality, they turn out to be 24 bit and not 8 bit. I have two scripts I've tried posted below this. Any ideas why it isn't compressing to 8 bit? Also, is there a reason why it only compresses to 24 bit? Thank you to any replys.

Here are a couple scripts I've tried:

im = imread(input_filename);

map = gray(256);

imwrite(im,map,output_filename);

This results in a 24 bit image.

Next:

im = imread(input_filename);

im8 = im2uint8(im);

imwrite(im8,output_filename);

This results in a completely black image.

[Luca Ferro]

does this help in any way? How to reduce the bit depth of the bmp image from 8-bit to n-bit, where n<8 andn>=1. Choose n=3. - MATLAB Answers - MATLAB Central (mathworks.com)

[Amanda F]

It helps clarify a few things, but I still can't seem to save an 8 bit file.

Answer 1

[Jonas]

you can stretch and compress it on your own, just as you want

im=imread('new.png');
im=rgb2gray(im);
class(im)

ans = 'uint8'

imshow(im);

im16=uint16(im)*2^8; % 16-8=8
imshow(im16)

im32=uint32(im)*2^24; % 32-8=24
imshow(im32)

% and similarly downwards

clearvars -except im32;
im16=uint16(im32/2^16);
imshow(im16)

im8=uint8(im16/2^8);
imshow(im8)

[Amanda F]

I tried to change it and then save like the example you showed, but it saved as a 24 bit image again. I don't understand. The image originally reads in with imread as a uint8, but then doesn't save as an 8 bit image.

Here is the script I tried:

im = imread(input_filename);

im32=uint32(im)*2^24; % 32-8=24

im16=uint16(im32/2^16);

im8=uint8(im16/2^8);

imwrite(im8,output_filename);

[Jonas]

first, you dont need each single step of compression, it should only show the principle

% if im is a grayscale image with 32 bit per pixel
im8=uint8(im/2^(32-8)); % from 32 bit to 8 bit

what is you output_filename? It sounds like the 8 bit grayscale image is saved is RGB image, which has 8 bit per pixel per channel, resulting into 24 bits per pixel

can you provide an exemplary 32 bit image please. Maybe our grayscale appears gray but is an RGB image. e.g. add an additional rgb2gray( ) before saving. But I can't tell you more without having an exemplary image

[Amanda F]

Thank you very much for suggesting I check the image for what color type was. I checked the image info and found the image as "True color". All my images were captured with no color, but the images still rgb and not grayscale, so using the rgb2gray() was required to make it work.

Here is the functional script:

im = imread(input_filename);

im8gray = rgb2gray(im);

imwrite(im8gray,output_filename);

[Jonas]

btw, your initial rgb image is not 32 bit, but already 24bit (3x8 bit). so it is already uint8, that's why you do not need those divisions described above

im=imread('RefImage6.bmp');
size(im,3)

ans = 3

class(im)

ans = 'uint8'

im=rgb2gray(im);
class(im)

ans = 'uint8'