Hi Gabriela, thanks for your answer.
I would like to know Why did You use this filter and not other for example GammaMap, BoxCar or others??? perhaps is Refineed Lee the best filter???
Hi camilomu
I dont’ know if Lee is better or not… I used Lee filter for to work with the mean and median … please take the look for this paper
Effects and performance of speckle noise reduction filters on active radar and sar images.pdf (699.7 KB)
Hi all
Below: Composition multitemporal in RGB for Change Detection, with ALOS PALSAR 1 images level 1.5, processed with Map Ready
HI all
Screenshot with composition multitemporal in RGB and log ratio bands, for an area in Venezuela, the images are ALOS PALSAR 1, in levels 1.5 and 1.1 for change detection, using amplitude and intensity, please take look this, the images 1 and 2 (left to right) are bands extracted from level 1.5 using amplitude; and bands 3 and 4 are levels 1.1 using intensity
1 Like
Hi all
I posted an screenshot with composition multitemporal in RGB, now I see radiometric distortions for this I maked different processings for ALOS PALSAR 1 images level 1.1…
In the top left the image was processed using correction for radiometric distortions (Terrain flattening), filter, deskew and terrain correction plus stacking.
In the top right image was processed using correction radiometric, filter, deskew, terrain correction plus stacking.
Below the image was processed using correction radiometric, multilook, filter, deskew, terrain correction plus stacking.
My question is: Why I not have better visualization with the terrain flattening?
there were some issues with the terrain flattening module as reported here:
I don’t know if it is resolved yet but I can remember that it worked better in earlier versions.
1 Like
So, I believe that continued the problem with TF in SNAP for version 5.0!!! 
Yes there are some issues with it. For 6.0 we’re planning to implement DEM oversampling that should improve the results and further improvements are planned for later versions.
Hi all
Someone knows which are the advantages and disadvantages of work with amplitude or intensity in change detection???
Thanks in advance
I would say it doesn’t really matter. As intensity is A² the contrasts are higher in the intensity image. That means dark areas become darker and therefore less comparable.
What I would recommend however, is calibrating intensity to Sigma0. Comparing intensities (or amplitudes) of different dates makes little sense because they don’t show the same backscatter. Only calibrated values can mathematically or statistically be compared.
Thanks ABraun!!!
Hi again
I have an new question about change detection, I applied three methods:
1.- Composition multitemporal in RGB ( combining the bands in Slave-Master-Slave for RGB), where the elements in purple color represents backscattering increasing, in green color backscattering decreasing and no change in black color.
2.- Segmentation, using the mean and standard deviation, where the mean more or less the standard deviation represent the change detection (in red color)
3.- Log ratio (in yellow color) based in likelihood ratio,
My question is: how I can to analyse the methods???
Regards and thanks for advance… 
Profile plot of an interest area for each bands of change detection,
Combination multitemporal in RGB
Log ratio
Segmentation
Greetings everyone!
Forgive me for my rather dumb question but I have to be sure in order to interpret correctly my results.
When you do change detection, you calculate band ratio within coregisterred products. Since there is Master and slave, is the ratio “Master image” / “Slave image” ?
It is not a simple ratio, but on the log likelihood, which is related to that. Please see here: Change detection. Which algorithm is implemented?
I did a simple ratio using bandmath and got the same values as the Change detection module, except for extreme values.
But I do not know if the extreme values are left apart by defining a threshold window in the Change detection workflow or it is because the module has a probabilistic approach and consider some values are unlikely to happen so it corrects them to the most likely value.
I’m not sure about that either, sorry 
Maybe someone can clarify.
Hi Hasina
This module have a probabilistic approach, ratio is based in Log Likelihood!
lveci stated that it is based on log likelihood: