Metadata-Version: 2.1
Name: watroo
Version: 0.0.4
Summary: A trous wavelet transform and utilities
Author-email: Frederic Auchere <frederic.auchere@gmail.com>
Project-URL: Homepage, https://git.ias.u-psud.fr/fauchere/wavelets
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE.rst
Requires-Dist: numpy >=1.20.3
Requires-Dist: opencv-python >=4.5.1
Requires-Dist: scipy >=1.7.2
Requires-Dist: numexpr >=2.8.1
Requires-Dist: tqdm >=4.64

# WATROO

Implements the _à trous_ wavelet transform and associated tools: denoising, enhancement, etc.

# Contents

[Installation](#installation)\
[_A trous_ transform](#a-trous-transform)\
[Scaling functions](#scaling-functions)\
[WOW! (Wavelets Optimized Whitening)](#wow-wavelets-optimized-whitening)\
[References](#references)

## Installation

Within the active environment

```sh
pip install .
```

## _À trous_ transform

`ATrousTransform` implements a dyadic 'à-trous' transform

## Scaling functions

### Triangle

### B3 spline

## Examples

### Denoise an image 

```python
import numpy as np
from watroo import AtrousTransform, Triangle

denoise_sigma = [5, 3]
transform = AtrousTransform(Triangle)
img = np.random.normal(size=(512, 512))
coefficients = transform(img, len(denoise_sigma))
# coefficients.data is an ndarray that contains the coefficients proper
coefficients.denoise(denoise_sigma)
# coeffcients accepts numpy operations
denoised = np.sum(coefficients, axis=0)
# which is equivalent to
denoised = coefficients.data.sum(axis=0)
```

The same result cam be obtained using the *denoise* convenience function

```python
from watroo import Triangle, denoise

img = np.random.normal(size=(512, 512))
denoise_sigma = [5, 3]
denoised = denoise(img, denoise_sigma, Triangle)
```

### Extract significant coefficients at a given scale

```python
# return a ndarray containing the 3-sigma significance of coefficients
# at scale 2 with hard thresholding
s = coefficients.significance(3, 2, soft_threshold=False)
```

### Compute the standard deviation of Gaussian white noise

```python
# compute 10 scales of the 2D B3spline
w = B3spline(2)
w.compute_noise_weights(10)
```

This returns a 1-D `ndarray` containing the normalization
used to estimate the significance of coefficients.

## WOW! (Wavelets Optimized Whitening)

```python
from watroo import wow
# read in your image here (must be floating point)
# ...
```

Standard enhancement:

```python
wow_image, _ = wow(image)
```

'Bilateral' version, slower but better:

```python
    wow_image, _ = wow(image, bilateral=1)
```

Denoised bilateral enhancement (best results):

```python
wow_image, _ = wow(image, bilateral=1, denoise_coefficients=[5, 2])
```

## References

* Starck, J.-L. & Murtagh, F. 2002, Handbook of Astronomical Data Analysis, Springer-Verlag, doi:[10.1007/978-3-540-33025-7](https://doi.org/10.1007/978-3-540-33025-7)
* Auchère, F., Soubrié, E., Pelouze, G., Buchlin, É. 2022, Image Enhancement With Wavelets Optimized Whitening, A&A, 670, id.A66, doi:[10.1051/0004-6361/202245345](https://ui.adsabs.harvard.edu/link_gateway/2023A&A...670A..66A/doi:10.1051/0004-6361/202245345)
