File Exchange

image thumbnail

MTV: modified total variation model for image noise removal

version 1.0 (149 KB) by

a modified total variation model is proposed for image noise removal



View License

Y. Wang, et al, MTV: modified total variation model for image noise removal, IEE Electronics Letters, vol.47, no.10, pp.592-594, 2011

in this MTV, the diffusion is along the edge direction of the original noisy image(maybe its smoothed version using a Gaussian filter of small scale),it can preserve edge and suppress staircase very well.

this code includes the TV, Perona-Malik, Y-K forth order PDE, and the proposed MTV model.

Comments and Ratings (8)

Diana Rahim

Hi Wang,
I'm new in matlab
How can I run your model?

Image Analyst

Image Analyst (view profile)

Can you change your screenshot to show what your code does? Like show both original and filtered images.

Wang YQ

Wang YQ (view profile)

Hi, Tiago ,

thank you for your interest in my work!

you can replace the "orgname" with '2', and specify the path of the file.

if your 2.bmp is a noisy image, one need to specify a noise-free image for reference(to calculate the PSNR, MSSIM), you can specify this noise-free image in the code
"if norn == 1,
    orgname = 'lena';%%%---------here is the noise-free reference image

hope helps to you!

Good luck!


Tiago (view profile)

Hi Wang,

This seems a nice work, and maybe something that can help me a lot.

However, I am new to matlab, and perhaps you may help me.

I have this '2.bmp' picture, and I would like to run it.

How can I use your model?

Wang YQ

Wang YQ (view profile)

I will publish the source codes of my other works on PDE,active contour and cardiac image analysis one after another, please pay attention! Thanks a lot!

Wang YQ

Wang YQ (view profile)

Dear BMV,

thank you for your interest in my work!

in the getMTV function, ac2 = ac*ac, just ac2 is used,and ac is also designed as an input arg. At the initial stage when I built the program, I think ac may be helpful for future, it is not used at present.

in addition, the MTV can be extended for color image, texture image, and even using the nonlocal strategy.


bmv (view profile)

mtv = getMTV(I,f,a,b,c,ac2,ac,lambda);
ac - unused. Why?

Wang YQ

Wang YQ (view profile)

the MTV model is formulated as the EQ.(4) in this Letter, it can also be interpreted as an energy minimization in Eq.(3).

although the paper has been pulished, after thinking for a period of time, I raised the following question:

In fact, it is easy to find the solution of Eq.(3) since the original noisy image is ready and it is not necessary to solve the PDE in Eq(4). On the other hand, the Eq.(4) works very well to find the final result. it seems there is a paradox.

I need someone to comment this paradox, ^-^, thank you !

MATLAB Release
MATLAB 6.5.1 (R13SP1)

Download apps, toolboxes, and other File Exchange content using Add-On Explorer in MATLAB.

» Watch video