dip-functionalization-matlab

Useful MATLAB Functions for Digital Image Processing

Digital Image Processing with MATLAB: A Quick Review

---

---

Basic intensity transformations

1. Linear Equation

s = a . r + b

• a = contrast, b = brightness
• For b > 0 brighter image
• For b < 0 darker image
• For |a| > 1 higher contrast
• For |a| < 1 lower contrast
• e.g. s = 2r
• Pixels values between 0 to L/a are mapped to 0 to L-1
• all pixel values greater than
• L/2, will be equal to L-1 (max value e.g. 255) (higher contrast)

---

2. Negative image

s = -r + 255

• For |a| > 1 higher contrast
• For |a| < 1 lower contrast

---

3. Power Equation

S = C . r^(gamma)

- if gamma < 1:
     • Higher contrast for low pixel values
     • Lower contrast for high pixel values
     • mapping is weighted toward higher output (brighter image)
     • e.g. s = 48 . r ^ (0.3) ----> brighter output

- elseif gamma > 1:
    • Higher contrast for high pixel values
    • Lower contrast for low pixel values
    • mapping is weighted toward lower output (darker image) 
    • e.g. s = 0.0004 . r ^ (2) ----> darker output

---

4. Log Equation

s = c . log2 ( r + 1 ) ---> greater c ~~~ higher max (brighter image)

---

5. Thresholding Equation

s = 1 / ( 1 + (m/r)^E ) ---> HIGH PASS EQUATION

• A simple tool used for image segmentation.
• The "threshold" is {m} and {E} controls the "slope" of the function.

% High-pass:
    % • low intensities are removed
    % • High intensities are highlighted
 % if E = 15, betweem 90, 150, 190 -----> m=190 has higher contrast and
 % shows more details

`s = 1 / ( 1 + (r/m)^E )` ---> LOW PASS EQUATION

- • High intensities are removed
- • Low intensities are highlighted

---

6. Histogram Equation

• h(r_k) = n_k

• imadjust used for Hist method

• g1 = im2uint8(mat2gray((1./(1+(m./f).^E1))));

• histogram equalization

f=imread(‘pelvis.tif’);
g=histeq(f);     % ---> Histogram equalization MATALB command
figure, imshow(g);
figure, imhist(g);

7. Sacling

$$ \text{scaledImage} = (\text{imgPixelValue} - \text{oldMin}) \times \left( \frac{\text{newMax} - \text{newMin}}{\text{oldMax} - \text{oldMin}} \right) $$

---

---

Some useful MATLAB functions (1):

% Example input image
I = double([0 0.5; 1.0 1.25]);

% Convert to different formats
J1 = im2uint8(I);            % 8-bit unsigned integer
J2 = im2uint16(I);           % 16-bit unsigned integer
J3 = im2int16(I);            % 16-bit signed integer
J4 = im2single(I);           % Single precision floating-point
J5 = im2double(I);           % Double precision floating-point
J6 = mat2gray(I);            % Normalized grayscale [0, 1]
J7 = im2uint8(mat2gray(I));  % Normalized grayscale to 8-bit

% Display results
disp('Original Image:');
disp(I);
disp('im2uint8:');
disp(J1);
disp('im2uint16:');
disp(J2);
disp('im2int16:');
disp(J3);
disp('im2single:');
disp(J4);
disp('im2double:');
disp(J5);
disp('mat2gray:');
disp(J6);
disp('im2uint8(mat2gray):');
disp(J7);

Function	DICOM Images	Other Images
im2uint8(I)	Not recommended	Good for normalized or low-dynamic-range data
im2uint16(I)	Recommended	Good for high-dynamic-range data
im2int16(I)	Rarely needed	Use if signed data representation is required
im2single(I)	Good for computations	Good for computations
im2double(I)	Good for precision-sensitive tasks	Overkill unless high precision is needed
mat2gray(I)	Useful for visualization	Useful for visualization
im2uint8(mat2gray(I))	Good for visualization	Useful for saving/visualization in 8-bit

---

imadjust function: for Linear and gamma correction

J = imadjust(I,[low-in high-in],[low-out high-out],gamma);

• [low_in high_in]

• The portion of your input intensity range that you care about.

• Any pixel in I with value ≤ low_in will be mapped to the lowest output level.

• Any pixel ≥ high_in will go to the highest output level.

• Everything in between is remapped (see formula below).

• [low_out high_out]

• The target range in the output image J.

• Pixels that fell below or at low_in become low_out.

• Pixels at or above high_in become high_out.

• In–between values are interpolated (and optionally gamma‐corrected).

% gamma < 1 -----> “brightens” mid‐tones (concave).
% gamma = 1 -----> straight linear mapping.
% gamma > 1 -----> "darkens” mid‐tones (convex).

`y = ((x – low_in) / (high_in – low_in))^gamma;`   % stretch & gamma‐correction
`y = y * (high_out – low_out) + low_out;`          % scale & shift
`y is clipped to [0,1]`

- NOTE for imadjust:
- The output image will be scaled 
- between 0-1
- So, it should be scaled to 0-255 again by:
- g=im2uint8(f)

-  f1=imadjust(f, stretchlim(f), [], 1); % [] --> default [0 1],
-  stretchlim(f) --> automatically finds the best Min and Max
-  the default limits [0.01 0.99], saturating the upper 1% and the lower 1%.
-  [0.02 0.97] --> saturate the upper 3% and the lower 2%.
-  g=im2uint8(f1);

---

imcomplement function: Making negative

output = imcomplement(input);

d=dicomread('1-200.dcm');
f=im2uint8(mat2gray(d));
g=imcomplement(f);

---

---

---

References:

1. Ghadiri, H. (2025). Digital Imaging (DIP in Medical Imaging course presentation, Department of Medical Physics and Biomedical Engineering, Tehran University of Medical Sciences, Tehran, Iran).
2. Russ, J.C. (1995) The image processing handbook. 2. ed. Boca Raton, Fla.: CRC Press.
3. Gonzalez, R.C. and Woods, R.E. (2018) Digital image processing. New York, NY: Pearson.
4. ‘Image Processing ToolboxUser’s Guide’.