Suggestion to implement a simple RESTFul microservice to retraive pixel value from GeoTiff

I am a student in Science in Computer Science (Master degree) and I am making a "simple" microservice based on GeoTiFF images and Spring Boot (Java).

What I would like to do is:

1. download the high resolution (10-m pixel) DEM Italy (i.e., TINITALY/01 ), divided into tiles, where each tile is a GeoTiff image.

"This DEM was obtained starting from separate DEMs of single administrative regions of Italy. The DEM is freely available as a 10 m-cell size grid (in GeoTIFF format), in the UTM WGS 84 zone 32 projection system (see the link download)".

http://tinitaly.pi.ingv.it/Download_Area1_1.html

2. save the images in a database (?)

3. develop a REST service in Spring Boot where, given a coordinate point (lat/long) and an around (of the pixel) returns an array of the DEM values

Questions (Thanks):

1. I have read that there are several approaches to save GeoTiff images such as Postgresql's PostGIS, mongoDB, GeoServer, etc. Honestly, for the application I need to make, I have not figured out which component is best for store GeoTiff.

2. I have read that several developers recommend, in order to work in the cloud, to transform GeoTIFF to Cloud Optimized GeoTIFF (COG). I am still studying the documentation but do you think it is a required step for the service I want to implement?

I worked with Postgres+POSTGIS and I think it works amazing if you just want to fetch values of cells - even spatial analysis will work well but maybe a bit slow for very large Raster dataset.

Here, how you can easily get the value of a cell using ST_Value: https://postgis.net/docs/RT_ST_Value.html as you see here you can use location or image space row, column to get the data. ST_Neighbourhood could be also useful in case you want to get values from nearby pixels: https://postgis.net/docs/RT_ST_Neighborhood.html

In terms of performance, I had worked with 30m*30m DEM of Victoria, Australia and it works really well using PostGIS - even for more advanced queries that needs spatial join the performance was acceptable.

I would recommend using PostGIS as I am pretty sure it will work well. But I am not sure about the second question - I didn't have experience with COG.

These are some sample queries that might be helpful. Here, I am using a point in WGS84 (SRID:4326) to get data from my raster data which is in SRID 7855 (so I need to transform the coordinates). My raster table is victoria_raster and the point coordinates in (latitude: -37.8136, longitude: 144.9631).

getting the pixel value (defined by geometry):

select rid, ST_WorldToRasterCoordX(raster.rast,
				ST_Transform(
					ST_Point(144.9631, -37.8136, 4326),
					7855
				)) as pix_value
from victoria_raster as raster
WHERE ST_Intersects(raster.rast,ST_Transform(
					ST_Point(144.9631, -37.8136, 4326),
					7855));

getting the values around the pixel (defined by geometry) - 3 by 3 as we only consider cols +- 1 and rows +-1 in the ST_Neighbourhood function:

-- 3 by 3 pixels - around the point
with input_point as (
	select ST_Transform(
					ST_Point(144.9631, -37.8136, 4326),
					7855) as geom
),
row_col as (
	select ST_WorldToRasterCoordX(raster.rast,
				input_point.geom) as c_number,
			ST_WorldToRasterCoordY(raster.rast,
				input_point.geom) as r_number,
			input_point.geom as geom
	from victoria_raster as raster, input_point
	WHERE ST_Intersects(raster.rast,input_point.geom)
)
select ST_Neighborhood(raster.rast, row_col.c_number, row_col.r_number, 1,1)
from row_col, victoria_raster as raster
WHERE ST_Intersects(raster.rast,row_col.geom)

So, I would recommend using database to handle the raster dataset - Postgres+PostGIS. And using queries in your REST webservice to fetch the values.