Pixel shift between Prisma L2D and Sentinel 2 L2A / Google Satellite Hybrid

[tazrart]

Dear all,

I wanted to identify a few objects on a Google Satellite Hybrid background displayed under QGIS via Search QMS , for example: photovoltaic canopies from car parks.

‌

The digitized contour (red outline) on the Google Satellite Hybrid background overlays correctly on the Sentinel 2 L2A (True-color R,G,B) image,

‌

However, on the Prisma image (True-color: B 33, B 23,B 11 ) I feel like there is a pixel shift…

‌

‌

Have you noticed this difference in your studies?

Below are the steps I did:

System Ubuntu, QGIS v3.22.3, Enmap-box v3.9

1. I uploaded a Prisma L2D image in the official site according to this setting:

‌

2. I converted the Prisma image L2D from .he5 format to .tif format via the Enmap-box module (project => add product => Prisma L2D)

‌

3. I converted the Sentinel 2 L2A image via the Enmap-box module (project => add product => Sentinel 2 L2A)

Note: The original coordinate system of the PRISMA and Sentinel 2 image is 32631 (EPSG), so I used the same system whether for the QGIS project or for the digitization layer (shapefile).

Thank you in advance for your response

[Andreas Janz]

Yes, subpixel shifts between different sensor products is a reality.

We also want to have an correction algorithm in the future, see #847.
@Daniel Scheffler and @Benjamin Jakimow , if I remember correctly, there is a third party plugin for AROSICS in development, right? What is the status, any news on that?

[Daniel Scheffler]

Yes, here is the repository. In QGIS, it can be installed via the plugin manager. It is called “Coregistration” there and implements some basic functionality of AROSICS for both global and local co-registration. So far as I know, it works quite well. But I can also recommend to use AROSICS directly if you need some more features.

‌

[Andreas Janz]

Thanks Daniel,
here it is:

‌

[Andreas Janz]

@tazrart if you try it out, please let us know if you like it as is.

[tazrart]

Thank you @Daniel Scheffler

@Daniel Scheffler

and @Andreas Janz for confirming this information and thank you for the proposal of the "Co-registration" plugin. So I installed this plugin under QGIS.

‌

• First, I prepared the reference image, I recorded a single band of Sentinel 2 and then I cut it according to the Prisma image frame. So both Sentinel 2 and Prisma images have the same extent:

‌

‌

• I did a first test with the "Automated local co-registration" module of this plugin :

‌

The reference image: one band of the sentinel image (already cut with the same Prisma frame).

The traget image: prisma L2D image

The other settings, I didn't change anything, I kept it by default.

‌

‌

‌

‌

• Unfortunately, I have the following errors:

‌

‌

Have you ever used this plugin? Is it an error due to the settings or is it a computer bug? are there application examples to better familiarize with this plugin?

Thanks again ,

[Daniel Scheffler]

This is an issue due to an incomplete installation. Please refer to the installation instructions in this repository. Note that I am not a maintainer of this QGIS plugin, so in case of installation issues, file an issue there or use AROSICS directly.

[Andreas Janz]

Have you ever used this plugin? Is it an error due to the settings or is it a computer bug? are there application examples to better familiarize with this plugin?

I never used it. As @Daniel Scheffler mentioned, the error is probably due to an incomplete installation. Have you read the doc?
https://github.com/SMByC/Coregistration-Qgis-processing#installation

[Andreas Janz]

If you have still problems with the installation or the plugin usage, please create a new issue in the issue tracker of the plugin: https://github.com/SMByC/Coregistration-Qgis-processing/issues/new

[tazrart]

Thanks for your feedback and directions. To be continued !

[Agustin Lobo]

@Andreas Janz:

Yes, subpixel shifts between different sensor products is a reality.

What is being reported here and actually find myself in my PRISMA images is not sub-pixel at all, it is several pixels. Does AROSICS work in such a case?

@tazrat any luck in your case?

BTW does enyone know if something like a PRISMA users forum exist?

[Daniel Scheffler]

Does AROSICS work in such a case?

Yes, it does. But by default, it treats shifts above 5 pixels as invalid (which can be adjusted).

[tazrart]

@Agustin Lobo Indeed, there are several pixels that are offset. I tested the Co-registration plugin, but it corrects only a few pixels and not all the pixels ot the image. I haven't found better... Surprisingly, I haven't found in the published articles (using this product) this type of correction!!

[Agustin Lobo]

@tazrat The most worrying is not the lack of published correction, but the lack of any mention of the problem in the articles using PRISMA. This is why I thought the problem was particular to my images. Is not there any forum dealing with PRISMA products?

@Daniel Scheffler could you perhaps contemplate an example of applying your software to fix the problem in PRISMA imagery?