Retina Preproccessing

Retina Preprocessing with CLAHE

Project Description

This project focuses on enhancing images using CLAHE (Contrast Limited Adaptive Histogram Equalization) to improve contrast and clarity. The preprocessing technique is especially useful for medical imaging, where clear visibility of features such as blood vessels or hemorrhages is crucial. CLAHE is applied to specified color channels to enhance features while maintaining image quality.

What Problem Does It Solve?

CLAHE preprocessing addresses the problem of insufficient contrast in images. This is particularly important for applications where detail and feature visibility are critical. By enhancing contrast adaptively, CLAHE helps in better feature extraction and analysis.

Features

Ability to appply CLAHE to different color channels (Red, Green, Blue).

Ability to resize images if needed.

Ability to save processed images to specified directories.

CLAHE clip limit input for easier fine tuning.

CLAHE tile grid size input.

Ability to display the input and output images if wanted.

Dependencies

OpenCV: For image processing and applying CLAHE.

NumPy: For numerical operations.

Matplotlib: For visualizing results. Features

Install instructions

Just download the file it's stand alone

But:

You need to install these libraries:

pip install opencv-python

python -m pip install -U matplotlib

pip install numpy

Use instructions

param filenameInput: Path to the input image file.

param filenameOutput: Path to save the output image.

param clipLimit: CLAHE clip limit. Float or int (Note: I have found 5.7-5.5 most successful)

param tileSize: CLAHE tile grid size. (Note: I have found 8,8 most successful)

param saveImages: Whether to save the output image. bool: True or False

param showImages: Whether to display the input and output images. bool: True or False

param channels_to_apply: List of channels to apply CLAHE ('R', 'G', 'B'). #any combination is possible as well as only one (Note: I have found RBG most successful but you can choose whichever you prefer)

param resize_dim: New dimensions for resizing the image (width, height). (Note: you can also put None if unneeded)

For example:

process_image_with_clahe(
    filenameInput="file path",
    filenameOutput="folder you want it in path\\ name of output",
    clipLimit=5.7,
    tileSize=(8, 8),
    saveImages=True,
    showImages=False,
    channels_to_apply=['R', 'G', 'B'],
    resize_dim= None
)

Patient with diabetic retinopathy

As you can see above ^^^ the blood vessels and hemorrhages are much more clear after application