Non Linear normalization of an image with IPP

[SergeyKostrov]

Hi everybody,

I wonder if IPP supports Non Linear normalization of an image? Please take a look at:

http://en.wikipedia.org/wiki/Normalization_(image_processing)

for more details.

Best regards,
Sergey

[Gennady_F_Intel · Accepted solution]

that's correct - there are no such normalization in IPP.

[SergeyKostrov]

Quoting Sergey Kostrov

Hi everybody,

I wonder if IPP supports Non Linear normalization of an image? Please take a look at:

http://en.wikipedia.org/wiki/Normalization_(image_processing)

for more details.

I'd like to make a newfeature request:

a function in Imageor Digital Signal Processing domainsfor a Non Linear Normalization of an image or a data set.

Best regards,
Sergey

[siddy]

Or, if you can generate the displacements by some means (Optical flow etc...), you can definitely apply the deformations using the Warp functionality (ippiWarp....). on a displaced coordinate list. This is the closest one can get to a non-linear deformation of an image in ipp, currently.

[SergeyKostrov]

Quoting siddy

Or, if you can generate the displacements by some means (Optical flow etc...), you can definitely apply the deformations using the Warp functionality (ippiWarp....). on a displaced coordinate list. This is the closest one can get to a non-linear deformation of an image in ipp, currently.

Thanks for the advise, Siddy! Actually, I don't need to transform an image geometrically. I need to reduce noise
in a very noisyimage and, at the same time,toenhance some details. To be honest, I've asked another person to evaluate
the equationfor Non-Linear Normalization ( from Wiki )and we have very strange results. I'll follow up with our R&D results
some time later...

[SergeyKostrov]

Quoting Sergey Kostrov

...I've asked another person to evaluate the equationfor Non-Linear Normalization ( from Wiki )and
we have very strange results. I'll follow up with our R&D results some time later...

Here is an example andfrom top to bottom are

Source set of "pixels", thenLinear Normalized and followed by Non-Linear Normalized:



Non-Linear Normalization looks like averaging( very similar to exponential averaging ). There is a question about
origins of the equation for Non-Linear Normalization.

[EDITED] Please see a Post #7. Unfortunately, an error was detected in the equation for Non-Linear Normalization.
It finally explainswhy we hadstrangeresults.

[SergeyKostrov]

Quoting Sergey Kostrov

...Non Linear normalization of an image? Please take a look at:

http://en.wikipedia.org/wiki/Normalization_(image_processing)

Attention: I'd like to to inform that the equation for Non-Linear Normalization of an imagehas an error.

[SergeyKostrov]

Here is an updatedexample andfrom top to bottom are

Source set of "pixels", thenLinear Normalized and followed by Non-Linear Normalized ( Alpha=0.5 & Beta=2.5 ):



I hope that it will be useful for a software developer or a researcher working on an image processing project.

Best regards,
Sergey

[SergeyKostrov]

Example of Linear andNon-Linear Normalization

Single-Precision Data Set (size: 1 x 16 )

>> Original Data <<

Range: Min = 1.000000 Max = 6.000000
1.000000 1.000000 2.000000 2.000000 1.000000 1.000000 4.000000 4.000000 1.000000 1.000000 6.000000 6.000000 1.000000 1.000000 1.000000 1.000000

>> Linear Normalized Data <<

New Range: Min = 0.000000 Max = 24.000000

0.000000 0.000000 4.800000 4.800000 0.000000 0.000000 14.400001 14.400001 0.000000 0.000000 24.000000 24.000000 0.000000 0.000000 0.000000 0.000000

>> Non-Linear Normalized Data <<

New Range: Min = 0.000000 Max = 24.000000

Test 1:
Params: Alpha = 0.032000 Beta = 2.500000
0.000000 0.000000 0.000004 0.000004 0.000000 0.000000 24.000000 24.000000 0.000000 0.000000 24.000000 24.000000 0.000000 0.000000 0.000000 0.000000

Test 2:
Params: Alpha = 0.064000 Beta = 2.500000
0.000000 0.000000 0.009708 0.009708 0.000000 0.000000 24.000000 24.000000 0.000000 0.000000 24.000000 24.000000 0.000000 0.000000 0.000000 0.000000

Test 3:
Params: Alpha = 0.125000 Beta = 2.500000
0.000147 0.000147 0.431669 0.431669 0.000147 0.000147 23.999853 23.999853 0.000147 0.000147 24.000000 24.000000 0.000147 0.000147 0.000147 0.000147

Test 4:
Params: Alpha = 0.250000 Beta = 2.500000
0.059343 0.059343 2.860870 2.860870 0.059343 0.059343 23.940657 23.940657 0.059343 0.059343 23.999981 23.999981 0.059343 0.059343 0.059343 0.059343

Test 5:
Params: Alpha = 0.500000 Beta = 2.500000
1.138221 1.138221 6.454594 6.454594 1.138221 1.138221 22.861778 22.861778 1.138221 1.138221 23.978134 23.978134 1.138221 1.138221 1.138221 1.138221

Test 6:
Params: Alpha = 0.750000 Beta = 2.500000
2.860870 2.860870 8.141848 8.141848 2.860870 2.860870 21.139130 21.139130 2.860870 2.860870 23.776417 23.776417 2.860870 2.860870 2.860870 2.860870

Test 7:
Params: Alpha = 1.000000 Beta = 2.500000
4.378212 4.378212 9.060976 9.060976 4.378212 4.378212 19.621788 19.621788 4.378212 4.378212 23.296507 23.296507 4.378212 4.378212 4.378212 4.378212

[EDITED] Added an example of Linear Normilized data