0% found this document useful (0 votes)
48 views4 pages

Assignment # 4: Image Denoising

This document discusses image denoising methods. It compares Gaussian filtering and non-local mean (NL-means) filtering. Gaussian filtering uses a linear symmetric kernel to convolve the image, while NL-means uses a weighted average of all pixels to reduce noise without blurring edges. The results show that NL-means more effectively retains image qualities and does not introduce noticeable artificial structures, as Gaussian filtering can.

Uploaded by

laxman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
48 views4 pages

Assignment # 4: Image Denoising

This document discusses image denoising methods. It compares Gaussian filtering and non-local mean (NL-means) filtering. Gaussian filtering uses a linear symmetric kernel to convolve the image, while NL-means uses a weighted average of all pixels to reduce noise without blurring edges. The results show that NL-means more effectively retains image qualities and does not introduce noticeable artificial structures, as Gaussian filtering can.

Uploaded by

laxman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Assignment # 4

Image denoising
18110093

Abstract
With such an increase in the number of images taken every day, the call for accurate and
pleasing images is increasing. The images captured by modern cameras are inevitably
degraded by noise. Work is required to reduce noise without losing image features (edges,
corners, and other sharp structures).

To break down an image we cans say that any image can be represented by;
v(i) = u(i) + n(i),
where v(i)is the observed value, u(i)is the “true” value and n(i) is the noise perturbation

Filtering methods
We have several methods to denoise an image but we are observing two of the most
fundamental methods; Gaussian filtering and non-local mean filtering.

● Gaussian filtering
This method involves convolution of the image by a linear symmetric kernel. The
Gaussian kernel has a standard deviation of h.

● ​NL-means algorithm
​Given a discrete noisy image v = {v(i) | i ∈ I}, where the family of weights {w(i, j)}j

The weights can be given as


The normalizing constant Z(i)

Results and conclusions


When we apply the gaussian blur(right), which is significantly unclear and implement
NLM filter(left), a somewhat clearer image.

The image here is a gaussian filtered on the Gaussian blurred image.


When we apply the gaussian blur(right), which is significantly unclear and implement
NLM filter(left), a somewhat clearer image.

The image here is a gaussian filtered on the Gaussian blurred image.


NLM filtered image(left) and image with salt and pepper noise(right)

Gaussian filter(left) and image with salt and pepper noise(right)

We find non-local means algorithm to be more efficient in terms of retaining image


qualities. NL-means method noise does not present any noticeable geometrical structures.

Reference:
https://www.iro.umontreal.ca/~mignotte/IFT6150/Articles/Buades-NonLocal.p
df
[1] L. Alvarez, P.-L. Lions, and J.-M. Morel. Image selective smoothing and edge detection
by nonlinear diffusion (ii). Journal of numerical analysis, 29:845–866, 1992.
[2] A. Buades, B. Coll, and J. Morel. On image denoising methods. Technical Report
2004-15, CMLA, 2004.

You might also like