How to change the gray level of a 8-bit gray level image to 7, 6, 5-bit gray level image using for loops?

Hello all,

Below is my code. So if I perform the operations outside of the nested for loop (meaning as a simple script where I input each value individually) I get the correct values for the pixels in the outImage. However, when I try to compute it within the nested for loop, the program doesn't stop executing, and when I click pause, it doesn't throw an error, rather it opens a new script with some chunk of code.

Does anybody have any idea of what do? I thought this would work.

Best,

close all;
clear;
clc;
 
% read the input image as input image
inImage = imread('rose_copy.jpg');
 
% get the shape of the image
[row,col] = size(inImage);
outImage = zeros(row,col);
 
n = 6; % the number of bits for the grayscale. Say I want it from 8-bit to 6-bit in this case
 
for y = 1:1:row
    for x = 1:1:col
        pixel = uint8(inImage(y,x));
        
        mask = uint8(256 - power(2,(8 - n)));
        
        pixel2 = uint8(bitand(pixel,mask));
        
        fin = bitsra(pixel2,8 - n);
        
        outImage(y,x) = uint8(fin);
    end
end
 
 
outImage = uint8(outImage);
imwrite(outImage,'Grayscale.jpg','jpg');

Answer 1

[Image Analyst]

To seqentially reduce the number of "used" bits in an 8 bit image you'd do

for k = 1 : 7
    grayImage = grayImage / 2;
end

[Jorge Vera Garcia]

I'm not sure if that is helpful, but would help me changing the gray scale of an image? If you could please expand on your exaplanation or let me know if my question seems weird.

Thanks.

[Image Analyst]

Not sure what you want. If you have an 8 bit image with values in the 0-255 range, then dividing it by 2 would now create an 8 bit image with values in the 0-127 range, which is a 7 bit range. Another division by 2 would get you into the rand 0-63 which is the 6 bit range, and so on. The gray levels definitely do change.

[Jorge Vera Garcia]

Oh got it, no, that's not what I am trying to do. I'm trying to perform image quantazation on a grayscale image. I have to go from 256 possible grays to 128 and 64. Not simply make it darker, if that makes sense.

My bad, I am new to image processing lol.

Let me know if you have any suggestions for that.

Thanks!

[Image Analyst]

So you want the full range of 0-255 but you just want it quantized into fewer gray levels. Did you try some functions called something like imquantize() and discretize() (not sure those names are right)?

[Jorge Vera Garcia]

So the whole thing with the assignment is to do it without those functions. I already figured something out. I was overcomplicating everything by using binary operations.

Thanks anyways for the prompt response.

[Image Analyst]

Oh - it wasn't marked as homework and you didn't originally say you had to write all your own code. I've tagged it as homework for you. But gladd you figured it out.

[Walter Roberson]

binary is not a bad way to approach it. You would probably get more accurate answers than successive division by 2, because in MATLAB divivision of uint8 rounds

x = uint8(1:10);
N = 2;
dec2bin(x)

ans = 10x4 char array
    '0001'
    '0010'
    '0011'
    '0100'
    '0101'
    '0110'
    '0111'
    '1000'
    '1001'
    '1010'

uint8(x)/2^N*2^N

ans = 1×10

0 4 4 4 4 8 8 8 8 12

uint8(round(double(x)/2^N)*2^N)

ans = 1×10

0 4 4 4 4 8 8 8 8 12

bitand(x, 255-2^N+1)

ans = 1×10

0 0 0 4 4 4 4 8 8 8

uint8(floor(double(x)/2^N)*2^N)

ans = 1×10

0 0 0 4 4 4 4 8 8 8

[Jorge Vera Garcia]

Thanks for the response. I actually implemented an alternative way that works perfectly.

Have a good rest of your day!