DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI...
Transcript of DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI...
![Page 1: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/1.jpg)
A NOVEL MEDICAL IMAGE
DENOISING ENTITY.
MAHESH MOHAN M RM.Tech student
Govt. Engineering College Trichur
OM NAMA SIVAYA
SHIVA KUDE KANANE…OmnamasivayaOm nama sivayaKUDE KANANE
DEVAOm nama sivaya
SHEEBA V SProfessor
Govt. Engineering College Trichur
![Page 2: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/2.jpg)
OVERVIEW
INTRODUCTION
WORK AT A GLANCE
PROBLEM FORMULATION`SPATIAL DOMAIN APPROACHES
TRANSFORM DOMAIN APPROACHES
PROPOSED WORK
CONCLUSION
![Page 3: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/3.jpg)
❖ The incorporated noise during image acquisition degrades the human interpretation, or computer-aided analysis of the images
❖ For a visual analysis of medical images, the clarity of details are
important.
❖ Two approaches to reduce noise in a medical image.
COMPEX ACQUISITION(Slower)
(High SNR)
INTRODUCTION
SIMPLE ACQUISITION(Faster)
(Low SNR)
1 2
![Page 4: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/4.jpg)
▪ Require long and repeated acquisition of the same subject to reduce noise and blur and to maintain a high SNR.
High SNR DTI (1 hour) High SNR HARDI (13 hours)
▪ To recover noisy and blurry image without lengthy repeated scans, post-processing of data plays a critical role .
MOTIVATION
![Page 5: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/5.jpg)
THESIS AT A GLANCE
❖ SPATIAL DOMAIN TECHNIQUES Bilateral filtering.NLm filtering.
❖ TRANSFORM DOMAIN TECHNIQUES DWT thresholding.Contourlet thresholding.
❖ PROPOSE A NOVEL ENTITY FOR MEDICAL IMAGE DENOISING.
France, Bourget du Lac – February 23, 2012
OMNAMASIVAYA
![Page 6: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/6.jpg)
Problem Formulation
❖ Noise in medical images can be generalised to Additive White Gaussian Noise (AWGN).
MEAN=0VARIANCE=0.05
MEAN=1.5VARIANCE=10
Original image
Noise image
ADDITION
DENOISING Entity
![Page 7: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/7.jpg)
SPATIAL DOMAIN METHODS?
Gaussian blur
Bilateral filter
NLm filtering
![Page 8: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/8.jpg)
Blur IN GAUSSIAN Comes from Averaging across Edges
*
*
*
input output
Same Gaussian kernel everywhere.
![Page 9: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/9.jpg)
Bilateral FilterNo Averaging across Edges
*
*
*
input output
The kernel shape depends on the image content.
[Aurich 95, Smith 97, Tomasi 98]
![Page 10: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/10.jpg)
Bilateral Filter on a Height Field
output
inputoutput
p
qWeigh I(q)By 2 * 0.4 = 0.8
r
Weigh I(r)By 48 * 0.4 = 19.2
![Page 11: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/11.jpg)
HOW TO ENHANCE PERFORMANCE OF BILATERAL FILTERING
p
❖ Bilateral very much depend on spatial intensities and thus abrupt noise values❖ Need to device preprocessing technique which removes the abrupt noise value retaining every edge information.
![Page 12: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/12.jpg)
NL-Means Filter (Buades 2005)
• Same goals: ‘Smooth within Similar Regions’
• KEY INSIGHT: Generalize, extend ‘Similarity’– Bilateral:
Averages neighbors with similar intensities;
– NL-Means: Averages neighbors with similar neighborhoods!
![Page 13: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/13.jpg)
NL-Means Method:Buades (2005)
• For each and
every pixel p:
Define a small, simple fixed size neighborhood;
Define vector Vp: a list of neighboring pixel values.
0.740.320.410.55………
Vp =
![Page 14: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/14.jpg)
‘Similar’ pixels p, q
🡪 SMALL vector distance;
NL-Means Method:Buades (2005)
|| Vp – Vq ||2
q
p
![Page 15: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/15.jpg)
‘Dissimilar’ pixels p, q
🡪 LARGE vector distance;
Filter with this.
NL-Means Method:Buades (2005)
|| Vp – Vq ||2
q
p
![Page 16: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/16.jpg)
HOW TO ENHANCE PERFORMANCE OF NLm FILTERING?
q
p
❖ NLm peformance very much depend on spatial intensities and thus abrupt noise values❖ Need to device preprocessing technique which removes the abrupt noise value retaining every edge information.
![Page 17: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/17.jpg)
Fig. 4.7 Comparing performance (a) Noisy image (σ = 40 ) (b) Mean filtering (c) Median filtering (d) Bilateral filtering (e) NLm filtering.
RESULT ANALYSIS OF SPATIALDOMAIN TECHNIQUES
Pls kude kananam
Comparing performance (a) Noisy image (σ = 40 ) (b) Mean filtering ,, (c) Gaussian filtering (d) Bilateral filtering (e) NLm filtering.
![Page 18: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/18.jpg)
… .Thresholding
TRANSFORM DOMAIN METHODS?
FORWARDTRANSFOR
M
REVERSETRANSFOR
M
Pls kude kananam
❖ DISCRETE COSINE TRANSFORM(DCT)❖ DISCRETE WAVELET TRANSFORM(DWT)❖ CONTOURLET TRANSFORM
![Page 19: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/19.jpg)
D.L. Donoho, I.M. Johnstone
Biometrika, vol. 81, no. 3, pp. 425-55, 1994.
“IDEAL SPATIAL ADAPTATION VIA WAVELET SHRINKAGE”
![Page 20: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/20.jpg)
WHY CONTOURLET IS SUPERIOR?
❖ WAVELET SATISFIES FIRST THREE WHILE CONTOURLET “ALL OF IT”.❖ BECAUSE CONTOURLET IS NOT A SEPERABLE TRANSFORM.
WHAT WE WISH IN A TRANSFORM?
MULTIRESOLUTIONLOCALIZATIONCRITICAL SAMPLINGDIRECTIONALITYANISOTROPY
![Page 21: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/21.jpg)
Fig. 4.7 Comparing performance (a) Noisy image (σ = 40 ) (b) Mean filtering (c) Median filtering (d) Bilateral filtering (e) NLm filtering.
RESULT ANALYSIS OF freQUENCYDOMAIN TECHNIQUES
Pls kude kananam
Comparing performance (a) Noisy image (σ = 40 ) (b) DCT based denoising (c) DWT based denoising (d) Contourlet based denoisng
![Page 22: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/22.jpg)
PROP0sed contributions
#1 Empirically formed a scaling factor for universal threshold for Visushrink.
#2 Introduced contourlet transform for denoising and empirically formed a similar scaling factor.
#3 Introduced a new entity for medical image denoising comprised of aforementioned contourlet thresholding as a preprocessing step to non-local mean denoising.
![Page 23: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/23.jpg)
CONTRIBUTION #1
THEORETICAL VALIDATIONUniversal threshold as derived by Donoho is hundred percent effective only when number of pixels in an image tends to infinity.
So a scaling parameter is deviced so that new threshold is Tw = λw *T. Where λw is
λw =3.944*10-11 S2 – 5.5285*10-6 *S+0.6022 and
![Page 24: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/24.jpg)
CONTRIBUTION #2
So a scaling parameter is deviced so that new threshold is Tc = λc *T. Where λcis
λc =3.944*10-11 S2 – 5.5285*10-6 *S+0.5522 and
THEORETICAL VALIDATIONFor image denoising, random noise will generate significant wavelet coefficients just like true edges, but is less likely to generate significant contourlet coefficients.
![Page 25: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/25.jpg)
EXPERIMENTAL RESULTSTABLE
![Page 26: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/26.jpg)
SIMULATION RESULTS #1 & #2
a b
c d
1 2 34 5
1-Noisy image(σ = 30) 2-Wavelet Univ threshold3-Contourlet Univ threshold4-Wavelet proposed5-Contourlet proposed
![Page 27: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/27.jpg)
CONTRIBUTION #3
THEORETICAL VALIDATION
Bilateral filtering –• real homogeneous gray levels corrupted by noise is
polluted significantly• fails to efficiently remove noise in regions of
homogeneous physical properties.
Contourlet denoising done before bilateral filtering,noise in homogeneous regions can be removed efficiently retaining the edge information as well as texture.
![Page 28: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/28.jpg)
THEORETICAL VALIDATION
NLM filter –• the calculation for similarity weights is
performed in a full-space of neighborhood. • Specifically, the accuracy of the similarity
weights will be affected by noise
CONTRIBUTION #3
Contourlet denoising done before NLm filtering,noise in neighbourhood can be removed efficiently retaining the edge information as well as texture.
![Page 29: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/29.jpg)
EXPERIMENTAL RESULTSTABLE
![Page 30: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/30.jpg)
SIMULATION RESULTS #3 & #4
a b
c d
1 2 34 5
1-Noisy image(σ = 30) 2-Bilateral filtering3-NLm denoising4-Proposed scheme BL5-Proposed scheme NLM
a b
c d
![Page 31: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/31.jpg)
PROPOSED MEDICALPROCESSING ENTITY
ADDITION
PROPOSED DENOISING Entity
NLm filteringProposedContourlet
thresholding
BACK
![Page 32: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/32.jpg)
PROCESSING TIME
s/m specification-4 GB RAM,2.30 Ghz processor.
Maximum exceeded time from Normal process = 0.2763 second
![Page 33: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/33.jpg)
❖ Improved the performance of Wavelet based thresholding.
❖ Introduced contourlet transform to image denoising and proved by simulation,proposed method is superior to wavelet transform.
❖ `By introducing proposed entity for common medical image denoisng techniques,performance can be significanly increased without significantly increasing time of processing.
CONclusion
![Page 34: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/34.jpg)
REFERENCESstogram?1. O. Christiansen, T. Lee, J. Lie, U. Sinha, and T. Chan. \Total
Variation Regularization of Matrix-Valued Images." International Journal of Biomedical Imaging, 2007(27432):11, December 2007.
2. S.Kalaivani Narayanan, and R.S.D.Wahidabanu. A View on Despeckling in Ultrasound Imaging. International Journal of Signal Processing, Image Processing and Pattern Recognition, 2009, 2(3):85-98.
3. H. Guo, J. E. Odegard, M.Lang, A. Gopinath, J. W. Selesnick. Wavelet based Speckle Reduction with Application to SAR based ATD/R. First International Conference on Image Processing 1994:75-79.
4. Rafael C. Gonzalez, Richard E. Woods,Digital Image processing using MATLAB. Second Edition, Mc Graw hill.
5. Lu Zhang, Jiaming Chen, Yuemin Zhu, Jianhua Luo. Comparisons of Several New Denoising Methods for Medical Images. IEEE, 2009:1-4.
6. C. Tomasi and R. Manduchi, "Bilateral filtering for gray and color images. " Proc Int Conf Computer Vision, pp. 839–846, 1998.
![Page 35: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/35.jpg)
REFERENCESstogram?7. D. L. Donoho and I. M. Johnstone, “Ideal spatial adaptation
via wavelet shrinkage,” Biometrika, 81, pp. 425–455, 1994.8. A. Buades, B. Coll, and J. M. Morel, “A review of image
denoising algorithms, with a new one,” Multiscale Modeling and Simulation (SIAM Interdisciplinary Journal), Vol. 4, No. 2, 2005, pp 490-530
9. M. N. Do and M. Vetterli, "Contourlets", in Beyond Wavelets, Academic Press, New York, 2003.Rafael C. Gonzalez, Richard E. Woods,Digital Image processing using MATLAB. Second Edition, Mc Graw hill.
10. J. B. Weaver, Y. Xu, D. M. Healy, and L. D. Cromwell, “Communications. Filtering noise from images with wavelet transforms,” Magnetic Resonance in Medicine, vol. 21, no. 2, pp. 288–295, 1991.
11. K.N. Chaudhury, D. Sage, and M. Unser, "Fast O(1) bilateral filtering using trigonometric range kernels," IEEE Trans. Image Processing, vol. 20, no. 11, 2011.
![Page 36: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/36.jpg)
REFERENCESstogram?8. D. L. Donoho and I. M. Johnstone, “Ideal spatial adaptation
via wavelet shrinkage,” Biometrika, 81, pp. 425–455, 1994.9. A. Buades, B. Coll, and J. M. Morel, “A review of image
denoising algorithms, with a new one,” Multiscale Modeling and Simulation (SIAM Interdisciplinary Journal), Vol. 4, No. 2, 2005, pp 490-530
10. M. N. Do and M. Vetterli, "Contourlets", in Beyond Wavelets, Academic Press, New York, 2003.Rafael C. Gonzalez, Richard E. Woods,Digital Image processing using MATLAB. Second Edition, Mc Graw hill.
11. J. B. Weaver, Y. Xu, D. M. Healy, and L. D. Cromwell, “Communications. Filtering noise from images with wavelet transforms,” Magnetic Resonance in Medicine, vol. 21, no. 2, pp. 288–295, 1991.
0m nama sivayaShivam shivakaram shantham shivathmanam shivothamam
Shiva marga pranetharam pranathasmi sada shivam…
![Page 37: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/37.jpg)
Equation of Gaussian filter
normalizedGaussian function
Same idea: weighted average of pixels.
0
1
![Page 38: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/38.jpg)
Gaussian FILTER
average
input
per-pixel multiplication
output
*
Gaussian blur
Bilateral filter
NLm filtering
![Page 39: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/39.jpg)
Fixing the Gaussian Blur”: the Bilateral Filter
space weight
not new
range weight
I
new
normalizationfactor
newSame idea: weighted average of pixels.
Box average
Gaussian blur
Bilateral filter
NLm filtering
![Page 40: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/40.jpg)
Bilateral Filter on a Height Field
output
inputoutput
p
qWeigh I(q)By 2 * 0.4 = 0.8
r
Weigh I(r)By 48 * 0.4 = 19.2
BACK
![Page 41: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/41.jpg)
wavelet transform?
Pls kude kananam
![Page 42: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/42.jpg)
Discrete wavelet transform?
Pls kude kananam
❖ A family of wavelets is then associated with the bandpass, and a family of scaling functions with the lowpass filters.
![Page 43: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/43.jpg)
Construction of 2d wavelet?
Pls kude kananam
LL LH HL HH
❖ Wavelet is a seperable transform.
![Page 44: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/44.jpg)
D.L. Donoho, I.M. Johnstone
Biometrika, vol. 81, no. 3, pp. 425-55, 1994.
“IDEAL SPATIAL ADAPTATION VIA WAVELET SHRINKAGE”
Hard thresholding Soft thresholding
BACK
![Page 45: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/45.jpg)
Construction of contourlet TRANSFORM.
BACK
![Page 46: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/46.jpg)
Test image
Wavelet Universal
Threshold
Contourlet
Universal
Threshold
Wavelet proposed
Threshold
Contourlet
proposed
threshold
σ =30
CT 19.2590 18.0033 22.7820 23.2653
MRI 20.8224 19.4452 23.7486 23.8390
σ =40
CT 16.1492 14.9014 20.2541 20.7107
MRI 18.2044 16.7866 21.4390 21.5036
σ =50
CT 13.8143 12.6572 18.4467 18.6677
MRI 16.2471 14.8722 19.7023 19.8100
EXPERIMENTAL RESULTS
BACK
![Page 47: DENOISING ENTITY A NOVEL MEDICAL IMAGE · 2020-05-19 · CT 16.1492 14.9014 20.2541 20.7107 MRI 18.2044 16.7866 21.4390 21.5036 σ =50 CT 13.8143 12.6572 18.4467 18.6677 MRI 16.2471](https://reader033.fdocuments.us/reader033/viewer/2022060406/5f0f56f07e708231d443abf9/html5/thumbnails/47.jpg)
EXPERIMENTAL RESULTSProposed entity Vs Simple bilateral and NLm filtering
BACK