General Steps for Sentinel-1 Data Preprocessing

See also this discussion:

1 Like

@mengdahl, @biocpu Just to clarify - thermal noise has not been removed from S1-A products (GRD or SLC). This can be checked by examining the tag in the annotation file (in the product annotation folder). This tag should be: false where false indicates that thermal noise correction has not been performed.

Products do include a thermal noise LUT to enable noise to be removed by the user but as already indicated, this is not been applied to the product (this LUT can be found in the noise annotation files).

2 Likes

Thank you Peter, that means we need to update our online documentation which indicates that the noise removal is performed by default:

https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar/product-types-processing-levels/level-1

Is there a specific reason why GRD thermal noise removal is not on my default?

1 Like

@mengdahl @peter.meadows Regarding to the Thermal Noise correction. So in general user have to remove the thermal noise, with S1TBX when the flag is set to false. On the other side nothing happens if you try to remove the thermal noise a second time. Am I wrong?

To sum up you can’t do something wrong if you always include the thermal noise reduction in your processing chain.

Actually it will give you an error that the thermal noise correction has already been applied and not let you continue.

@lveci
Yes that’s correct. Thank you!

Hello all,

I am doing my master on sea ice observation using a SENTINEL-1 EW GRD product for an area of Greenland. The problem i have is that the SENTINEL-1 EW_GRD product i have downloaded (2015 December) has some bright stripes along the image (I know that this problem has been raised before, i have seen the commends but i am bit confused).

I have read on many posts that this issue is due to thermal noise. Once i applied thermal noise removal, the bright stripes disappeared and dark pixels appeared instead (i have the images attached). The thermal noise removal has not been properly worked.

So, my question is, Does the image quality can be improved any further? or I have to accept that nothing else can be done on the image. So, I move on and post-process the SENTINEL 1 iproduct as is

I would appreciate any help

Thank you in advance

Ioannis

The second image looks much improved. I believe the thermal noise vectors in the product are intended for correction over areas with higher backscatter than ocean. Over ocean you may need to look at the average cross section and come up with your own normalization. In one image it may be easy to do but it may be more difficult to come up with a general solution that will work with any scene. If someone has a general correction that’s easy to apply, we could implement it in the toolbox.

Thank you for your quick reply

Ioannis

Hi there, I did these pre-processing for my Sentinel 1 data: Thermal noise removal–> Apply Orbit file --> Calibration to beta ) --> Radiometric Terrain flattening --> Range Doppler Terrain.
So I have been told that it is ok to skip “Speckle filtering”. And now I have been told by others that “Speckle filtering” is necessary. I was wondering if I can do it right now after Radiometric Terrain Flattering and Range Doppler Terrain Correction? DO you think doing Speckle filtering as the last step might harm my results?
Thanks for your reply in advance

1 Like

I would personally say that it strongly depends on your image. Speckle filtering can surely increase image quality and it is done best at an early stage so speckle is not falsely propagated in ongoing resampling processes (within terrain correction, for example). But if you are interested in smaller structures or image texture, speckle filtering might remove the information you are interested in.
So saying speckle filtering must be done is surely not correct, but it is advisable in many cases.
If you do it after terrain correction you have to notice that your image was already resampled to a new resolution and coordinate system. So your pixel values were already changed (and in case of bilinear resampling this involved a first smoothing). So the results might be different but it is advisable to compare.
Please also note that speckle filters perform different when applied on the db image.

3 Likes

Hello
Can I know how I can obtain the noisy image from the product
S1A_EW_GRDM_1SDH_20160218T113753_20160218T113837_010000_00EB48_F483
Thanks in advanced

In case anyone is interested in using snappy (SNAP Python module), I have written a script to pre-process Sentinel-1 GRD products using snappy https://github.com/wajuqi/Sentinel-1-preprocessing-using-Snappy .

The processing steps I used are:

  1. Apply orbit file
  2. GRD border noise removal (optional)
  3. Thermal noise removal
  4. Radiometric calibration
  5. Speckle filtering
  6. Terrain correction
  7. Subset
2 Likes

One correction, speckle-filtering should be done in radar-geometry, so before the Terrain Correction step.

Thanks! Corrected.

Dear @ABraun ,
According to your explanation, should we perform speckle filter before the radiometric calibration (before having dB image)?

Thank you!

@JunqianW, May I know the reason why do you apply the image subset at the last stage instead of the beginning to reduce the file size and computing resource?

yes, most speckle filters are designed for the power-scaled data

1 Like

@ABraun, so the procedure for Sentinel-1 pre-processing should be operated as follow sequence!?:

  1. Apply orbit file

  2. GRD border noise removal (optional)

  3. Thermal noise removal

  4. Speckle filtering

  5. Radiometric calibration

  6. Terrain correction

  7. Subset

You can perform the subset earlier, but yes.

1 Like