uint8 image solarization issue

Good afternoon I am trying to take a uint8 image and take the complement of all pixels with a grayscale less then 128 and again with all pixels with a grayscale greater than 128. This is an example of my attempt but I don't think it is working properly, if anyone could please give me a hand. I am supposed to create the two functions and then create a small test image to test the functions before applying them to an image but I don't know how to do that.

          b=imread('blocks.jpg');
          class(b)
          comp_light_pixels=b<128   %only complement the lighter pixels of b
          subplot(2,2,1)
          imshow(comp_light_pixels)    %show image with lighter pixels complemented
          comp_dark_pixels=b>128    %only complement the darker pixels
          subplot(2,2,2)
          imshow(comp_dark_pixels)   % show image with dark pixels complemented
          subplot(2,2,3)  
          imshow(b)   % show original image
          subplot(2,2,4)
          imshow(255-b)   % show completely complemented image

Answer 1

[Walter Roberson]

 comp_light_pixels = b;
 mask = comp_light_pixels < 128;
 comp_light_pixels(mask) = 255 - comp_light_pixels(mask);

[Aaron Connell]

this did not work unfortunately Sir. This is how I got it to work, but I thank you for the answer either way

the_image=imread('sample_image.png');   % read in the desired image
y=uint8(the_image);    % make sure image is uint8 before performing solarization

subplot(2,2,1)  
imshow(y)   
title('Original UINT8 Image')

subplot(2,2,2)
imshow(255-y)   
title('Completely Complemented Image')

mask = y < 128; %only complement the pixels with grayscales less than 128, or the darker pixels
convert_mask=uint8(mask);
image=convert_mask.*y + (1-convert_mask).*(255-y);
subplot(2,2,3)
imshow(image)
title('Darker Pixels Are Complemented')

mask1=y > 128;   %only complement the lighter pixels
convert_mask1=uint8(mask1);
image1=convert_mask1.*y + (1-convert_mask1).*(255-y);
subplot(2,2,4)
imshow(image1) 
title('Lighter Pixels Are Complemented')

%%Question 1 Part C: Use the solarization functions on the bergen image.
the_image=imread('bergen.jpg');   % read in the desired image
y=uint8(the_image);    % make sure image is uint8 before performing solarization

subplot(2,2,1)  
imshow(y)   
title('Original UINT8 Image')

subplot(2,2,2)
imshow(255-y)   
title('Completely Complemented Image')

mask2 = y < 128; %only complement the pixels with grayscales less than 128, or the darker pixels
convert_mask2=uint8(mask2);
image2=convert_mask2.*y + (1-convert_mask2).*(255-y);
subplot(2,2,3)
imshow(image2)
title('Darker Pixels Are Complemented')

mask3=y > 128;   %only complement the lighter pixels
convert_mask3=uint8(mask3);
image3=convert_mask3.*y + (1-convert_mask3).*(255-y);
subplot(2,2,4)
imshow(image3) 
title('Lighter Pixels Are Complemented')

[DGM]

Granted, it's not like "solarize" is really a well-defined transformation. That said, the given example still doesn't do what's probably expected.

% the image can only be uint8 class
% otherwise, everything will fail
x = uint8(0:255);

% only complement the pixels with grayscales less than 128, or the darker pixels
mask = x < 128;
convert_mask = uint8(mask);
y1 = convert_mask.*x + (1-convert_mask).*(255-x);

% only complement the lighter pixels
mask1 = x > 128;
convert_mask1 = uint8(mask1);
y2 = convert_mask1.*x + (1-convert_mask1).*(255-x);

% plot the curves
plot(x,[y1;y2])
xlim([0 255])
ylim([0 255])

As is common, it's a hard vee curve, but neither case is full swing. That does mean that the contrast is preserved on either side of 50% gray, but I've never seen such an implementation, and have no idea why that would be a desirable interpretation.

This is a far simpler way to do the work. It's a more typical curve, and it's not blindly dependent on the input image class.

% some image in any class
x0 = uint8(0:255);

% put it in a consistent and convenient class and scale
x = im2double(x0);

% interpolate
tf = [0 1 0]; % hard vee curve
%tf = [1 0 1]; % inverted hard vee
y = interp1([0 0.5 1],tf,x,'linear');

% cast the output if you want it in some particular class
y = im2uint8(y);

% plot the curve
plot(x0,y)
xlim([0 255])
ylim([0 255])

The output curve is full swing, as is more typical. The input image can be of any typical numeric class.

% some image in any class
inpict = imread('peppers.png');

% put it in a consistent and convenient class and scale
inpict = im2double(inpict);

% interpolate
tf = [0 1 0]; % hard vee curve
%tf = [1 0 1]; % inverted hard vee
outpict = interp1([0 0.5 1],tf,inpict,'linear');

% cast the output if you want it in some particular class
outpict = im2uint8(outpict);

% display the result
imshow(outpict)

Of course, there are existing tools to do this.

% some image in any class
inpict = imread('peppers.png');

% solarize it (MIMT solarize())
%outpict = solarize(inpict,'vee'); % simple vee curve
outpict = solarize(inpict); % a nonlinear curve

% display the result
imshow(outpict)