I’ve downloaded a Sentinel-2 product and looked at the RGB image in SNAP.
I have areas who suffer extremely from fading/cross-fading
Here is a small crop of an industrial area:
Do I have to perform some kind of processing? How can I get rid of this?
It’s not clear to me what you mean by “cross-fading” (misregistration?) and the image you have posted is a wee bit unclear for my tired old eyes. Do you have the Tile and Orbit number so I can check it from here, please?
Thanks in advance
S2 MPC Operations Manager
it is tile 32UMV (Date: 2016_8_23).
Here is another image:
By fading (german: Überblendung ) I mean that in the industrial area the white color seems to blur the image. The structure of the industrial area/the buldings is not shown very good. I marked some extrem areas.
I am new to the area of remote sensing. Sorry for using an unusual term.
I know that Google Earth data is not equal to Sentinel-2 data. However, this image shows some of the above region in an similar resolution. The image from the Sentinel-2 seems not to show the correct color.
It may be due to how the SNAP tool performs its assessment of the image colour. I know that for some images, due to a combination of sun angle and reflectance, some areas in the Tropics that have hightly reflective corrugated tin rooves (unpainted) show up as if they are saturated. You should also be aware that the Google image made up from VHR pixels, at a much smaller resolution than S2, and thus more discrimination and variance of colour is possible when compared to the S2 10m pixel spatial resolution in the visible Bands.
In the RGB view in SNAP, you can use the Colour manipulation (View > Tool Windows > Colour Manipulation) to vary the RGB bands.
I am downloading the Tile, but it is part of the old slice format (as opposed to the new single Tile delivery introduced recently), and is therefore part of a 6.9 GB download…
Ok. Thank you very much for your effort. Does the color manipulation really help? I wasn’t successful.
More important: Is there no way to get rid of this in an automated way (e.g., using a script)?
What also bothers me is that the images provided here do not suffer from such effects.
Example 1: http://www.esa.int/spaceinimages/Images/2016/02/Barcelona_from_Sentinel-2A
Example 2: http://www.esa.int/spaceinimages/Images/2016/11/Sentinel_sees_us
I performed a stretch (colour manipulation) on the RGB (bottom right of the image), and it can be made to show individual aircraft at Frankfurt airport (whereas the RGB that SNAP generated initially looked similar to yours).
I would also note that the Band 8 has a good representation of tones in grayscale (bottom left of the image).
I’ll keep investigating.
It’s generally a good practice to adjust manually the colour range and not let SNAP do its own automatic adjustment. Setting the saturation at 3000 digital counts (i.e. reflectance of 0.3) leads to satisfactory results for most scenes. Exceptions are the Sahara and snow-covered areas where the reflectance can reach 1. Be sure to put the same settings for all R, G, B bands.
I attach the image I generated with these settings, it looks pretty nice to me.
Hello Sebastian and Jan,
thank you very much for your effort and support. Indeed, the industrial area now looks quite good. However, the darker forest areas seem to suffer from the color manipulation in a way that the structure is not that clear as in the beginning. But I thing I have to live with this…Thanks, good job
I want to investigate products from different places all over the world in an automatic approach. Please correct me if I am wrong: If I want to get rid of this problem automatically, the best thing I can do is to set the saturation at 3000 (or maybe 2000) digital counts (for areas without heavy snow or Sahara) or is there anything more I can do? I think I will have a look if I can access the color manipulation of Snap using Python (using snappy/snap-python).
Is there any kind of documentation where I can find such rules (to set the saturation at 3000 digital counts) or is this just your experience-based knowledge?
Another question: What exactly do you think is the reason for this problem here?