Border artifacts in SNAP v9 but not SNAP v8

snap
#1

Hi,

I’ve noticed border artifacts appearing in a Sentinel-1 image, but only when using SNAP v9. Here are the comparisons:

2022-11-01-SNAP_8_vs_9.pptx
2022-11-01-SNAP_8_vs_9.pptx1280×720 172 KB

To reproduce:
Using this Sentinel-1 image

  • Filename: S1A_EW_GRDM_1SDH_20180313T181225_20180313T181325_021000_0240E1_8163

  • Date / Time: 2018-03-13 18:12:25
    process using this xml file
    process.xml (7.1 KB)

one time with SNAP v8 and one time with SNAP v9. I’ve also created a diff image, as well as diff images for bands 1, 2, and 4 (HH, HV, and angle).
Difference all bands:

br_8163_dB_rafaelm_difference
br_8163_dB_rafaelm_difference1300×815 117 KB

Difference HH:
br_8163_dB_rafaelm_1_difference
br_8163_dB_rafaelm_1_difference1300×815 88.3 KB

Difference HV:
br_8163_dB_rafaelm_2_difference
br_8163_dB_rafaelm_2_difference1300×815 92 KB

Difference Angle:
br_8163_dB_rafaelm_3_difference
br_8163_dB_rafaelm_3_difference1300×815 6.4 KB

Any ideas what might be causing the discrepancy? Thank you for your help!

#2

Maybe @jun_lu or @lveci could help you.

#3

A JIRA ticket ([SNAP-3579] - JIRA) has been created to track the issue. We will look into the problem and thank you for reporting the issue.

#4

The artifacts were caused by the range noise vectors used in thermal noise removal. The image below shows the thermal noise of the given EW product. The noise value for the bright area is about 1e6 while the noise value for the dark area is around 1e3. We believe the noise vectors of the product for the bright area are not correct and that lead to the artifacts in the noise corrected images. We have checked several other EW products and did not find noise value that high.

image
image804×760 28.6 KB