scatter-downsample 0.2.5

A module used for downsampling high resolution images using the scattering transform

Downsampling high resolution images using the scattering transform

Atlasing high resolution data,Scattering transform, Visualization of imges, Downsampling with multiple methods

Usage

python3 -m scatter_downsample -i IMAGE -n NUMBER OF DOWNSAMPLE [-s] [filtered standard deviation channels] [-d] [directions for gaussian] [-m] [methods]

or

import scatter_downsample in python3 environment

ScatterDown(img,ndown2,filtered_std=2,method=1,direction=2):

Input:

- img: input image
- ndown2: how many times to downsample by a factor of 2
- filtered_std: how many standard deviations maximum should not get filtered out (default 2)
- direction numbers: for gaussian downsampling (default 2)
- method: (default 1)
    - 2*2 mean and standard deviation (1)
    - 2*2 mean and standard deviation with directions (2)
    - Gaussian filtering (3)

Output:

- high dimensional image in ndarray form
- labels for each channel
- an example visualization that you save as a png file

Tests

installed by: python -m unittest tests/testcase.py

• an example image to test for the shape of output image and labels for method 1 (mean and std)

Versions

0.2.4

• Add options for Gaussian Downsampling
• Add Gaussian Flipping and Rotations methods
• Setup requirements.txt for auto download required modules

0.2.0

Add in updated methods (mean/std with directions and gaussian filtering), add command line interfaces with argparse