Hacker News new | past | comments | ask | show | jobs | submit login
The Musaicum EU-plus 10M resolution image of Europe (imagico.de)
138 points by hk__2 on Sept 3, 2023 | hide | past | favorite | 25 comments



Good article, good end result dataset.

Worth it for the discussion regarding optimal scales for aggregating pixel stats.


What is the total file size of this?


It’s roughly 5000 x 5000 km, so 250 gigapixels. With 24-bit color, that would be about 700 GB uncompressed. Lossless compression can probably reduce that by a factor of ten or more.


Which lossless compression would compress images like this by a factor of ten or more? Isn't lossless image compression ratio more like around 2:1?


It’s just a rough guess based on the redundancy in that image (limited range of colors, no hard edges). Lossless compression ratios can vary wildly depending on image contents.


What we are really getting at here is: where's the jpg!?


what's the highest resolution satelite data that is commercially available actually?


10cm, available from e.g. https://nearspacelabs.com/ (using stratospheric balloons), or in 2025 from https://albedo.com/ (using satellites).


worldview3 and pléiades neo have about 30cm per pixel (the images are sold by third-party vendors such as maxar, airbus)


How are Airbus and Maxar 3rd party vendors here? They own and operate the satellites themselves


This may be a little off topic, but does anyone know of high resolution public domain datasets that can be used as basis for the rendering of the Earth's surface? Normal, albedo, etc.? Do they even exist? I know Nasa has some but what I found was tiny (100k res images and such are really low res when we're talking about the entire planet).


Sentinel 2 images are freely accessible, paid by european citizens through the copernicus program. There's also landsat 8 images, of similar resolution (10m/pixel for visible, lower in infrared bands). The revisit time depends on the latitude, but roughly you have a new image of the whole earth every week.

Maybe that's too high res for your application?


https://scihub.copernicus.eu/

No clue if it's enough for your purposes, but I think this is the basis for the article


USA only and with caveats but the US Department of Agriculture operates the National Agricultural Imagery Program:

https://naip-usdaonline.hub.arcgis.com/


OpenStreetMap tiles maybe?


OpenStretMap does not have aerial imagery per se, several imagery layers are available only for tracing features off them: https://wiki.openstreetmap.org/wiki/Using_aerial_imagery#Ima...


I would suggest lower casing the "M" as it stands for "meters" . By reading the submission title (the article one is correct, I don't know why you changed it) I couldn't figure out which mega units it was referencing (pixels, bytes,?)


Lowercase to uppercase might have been an automatic change by HN too.


HN titles can be edited by the submitter after submission. The automatic normalization only happens on the initial submit.

It would be nice if HN would display a message after submitting that the submitter should check the changed title.


Then spell it out as "10-meter". Having a clear and understandable headline is better than saving a few characters.


Eh, this is something HN does automatically after submission. It also sometimes drops words.


Submitters can re-edit the title after the fact.


I saw this comment several hours after submission and it was already too late to edit the title. I copy/pasted the title from the blog, so I guess it’s HN who capitalized that letter.


I still don't quite understand what 10m means in this context. Sorry, I'm slow.


1 pixel = 10 meters

i'm slow too




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: