lisabug/guided-filter

Implementation of Guided Image Filtering.

Guided Filter

Simple Python implementation of paper:

K.He, J.Sun, and X.Tang. Guided Image Filtering. TPAMI'12.

Guided filter is both effective and efficient in a great variety of computer vision and computer graphics applications, including edge-aware smoothing, detail enhancement, HDR compression, image matting/feathering, dehazing, joint upsampling, etc.

Visualization of results

Reimplementation of Figure 2 in paper and visualization on gray image.

And visualization on color image for denoising.

Dependencies

This repository depends on Python modules listed below.

• numpy
• matplotlib
• opencv

If you want to run demo main.py by yourself, you have to install those modules in advance. On the other hand, core of guided filter only depends on numpy, and you can check it for more.

Usage

from core.filter import GuidedFilter
# load your image
GF = GuidedFilter(image, radius, eps)
GF.filter(filtering_input)

Demo

You can run demo for quick visualization on guided filter by

> python main.py

License

The MIT License 2017 (c) Lisabug

guided-filter's Issues

Erroneous covariance matrix computation

Hi!
I believe that you have forgotten to subtract mean when computing covariance matrix here.

I think you need to use something like this:

corrI = corrI - np.matmul(meanI_, meanI_.transpose(0, 2, 1))

to_8U Function problem

It is generating most of the images full Black.