IVHC (Fast image noise estimation)

This is an implementation of IVHC on Python and Matlab. See also IVHC. IVHC is a model to estimate Gaussian, signal-dependent, and processed noise in image and video signals. The estimation is based on the classification of intensity-variances of image patches in order to find homogeneous regions that best represent the noise.

Here is the block diagram of the intensity-variance homogeneity classification (IVHC) noise estimation.

Inputs:

• Noisy gray image
• Max polynomial regression degree

Outputs:

• Variance of noise in the Y channel (best representative)
• Degree of processed noise
• Noise level function

The repository includes:

• Matlab and Python implementation of IVHC.
• Matlab demo files to estimate AWGN, processed noise, and signal-dependent noise.
• Python demo files to estimate AWGN, processed noise, and signal-dependent noise.

Python

Getting Started

• demo.ipynb or (demo.py) is the easiest way to start. It shows an example of estimating three types of noise. AWGN, PPN, and PGN.

Python Installation

1. Install dependencies pip3 install package [numpy, scikit-mage, ...]

2. Run setup from the libs directory python3 setup.py install optional:

   • run "python3 setup.py build" and copy .so (linux) or .pyd (windows) file to the demos.py path
   • if you have python3.9 copy "ivhc.cpython-39m-x86_64-linux-gnu.so" (linux) or "ivhc.cp39-win_amd64.pyd"
   • if you have python3.10 copy "ivhc.cpython-310m-x86_64-linux-gnu.so" (linux) or "ivhc.cp310-win_amd64.pyd" (windows) to your demos.py path.

3. Run demos.py: python3 demos.py

Matlab (windows only)

• demo_awgn.m is the easiest way to start. It shows an example of estimating AWGN.

• demo_pgn.m PGN (signal-dependent) noise estimation.

• demo_ppn.m PPN processed noise estimation.

• demo_real.m non-synthetic (real) image noise.

• demo_compare_awgn.m compare AWGN with other method.

• demo_compare_ppn.m compare PPN with other method.