This project is developed in Python + OpenCV for implementing some techinques of image processing and feature extraction algorithms.
- Geometric Transformations(e.g. rotation, image stitching)
- Histogram Equalization
- Interpolation
- Image Adjusting
- Histogram Equalization
- Local Histogram Equalization(LHE)
- Box Filter
- Median Filter
- Noises(e.g. salt-and-pepper, gaussian)
- Sharpening, Blurring and Noise Removal
- Edge Detection(first-order difference, robert, sobel)
- Unsharp Masking
- Discrete Fourier Transform(DFT)
- Separable Filters
- Filtering
- Color Spaces(e.g. RGB, HSI)
- Quantization
- Clustering
- Discrete Wavelet Transform(DWT)
- Multi-level DWT
- Approximation & Prediction Residual Pyramids
- Denoising with Wavelet
- Estimate Geometry
- Corner Detection(with Harris algorithm)
- Python: 3.9
- OpenCV: 4.5
- Numpy: 1.22
- Pandas: 1.3
- Matplotlib: 3.4
- Scikit-image: 0.18
- PyWavelets: 1.2
- Clone the repository:
git clone https://github.com/mohsenMahmoodzadeh/image-processing-basic-methods.git
- Set up Git-LFS:
# install the package
apt-get install git-lfs
# initialize it on the project
git lfs install
# modify the git config not to download large files at first
git config --global filter.lfs.smudge "git-lfs smudge --skip -- %f"
git config --global filter.lfs.process "git-lfs filter-process --skip"
# pull the large files into the project
git lfs pull
- Create a virtual environment (to avoid conflicts):
virtualenv -p python3.9 fcv
# this may vary depending on your shell
. fcv/bin/activate
Install the dependencies:
pip install -r requirements.txt
If you want to work on jupyter notebook, you may need to set up a kernel on your virtual environment to make sure all your modules execute correctly.
python -m ipykernel install --name fcvkernel
# Now you get a kernel named `fcvkernel` in your jupyter notebook
-
Correct and improve the image registration algorithm in 01 Image Fundamentals.ipynb notebook.
-
Improve the algorithms in 02 Contrast Adjustment.ipynb and 03 Filter.ipynb
-
Apply OOP to algorithms and visualization plots to gain reusability.
Fixes and improvements are more than welcome, so raise an issue or send a PR!