Rasterio: select subset of georeferenced image by lop left coordinates and height, width

Question

I want to subset a 10m sentinel dataset using rasterio window modul. I have tried using the window = from_bounds(...) part in .read() however it requires for me to put in the left top right bottom coordinates.

I want to subset using only the top left coordinates and the height and width of subset in pixels.

Here is what I have tried so far

import rasterio 
from rasterio import plot # for plotting datasets, may not be of use
from rasterio.plot import show_hist # for showing histogram distribution of bands in a daatset, may not be of use
from rasterio.windows import Window , from_bounds, crop# for clipping out only a window from dataset use in part 2
import matplotlib.pyplot as plt #for plotting images
import numpy as np # for doing computations
import os  # to read file paths to the image data
from sklearn import cluster
%matplotlib inline

path_10 =  R'./GRANULE/L2A_T31UET_A025788_20200530T105134/IMG_DATA/R10m'
l_10 = os.listdir(path_10)
tostack_10 = l_10[1:5] #take bands B G R NIR for 10m data
stackpath_10 = [os.path.join(path_10,i) for i in tostack_10]
ds_10 = [rasterio.open( stackpath_10[i], driver='JP2OpenJPEG' ) for i in  range(len( stackpath_10 ))] # list of datasets    

bands_10= [] #array holding all bands as numpy arrays 

# window_= Window( 601200.0, 5773695.0, 700, 500)

for i in range( len(ds_10)):
    bands_10.append( np.array(ds_10[i].read(1 ) , dtype='int32' ) )
    ds_10[i].close()


bands_10 = np.array(bands_10 , dtype='int32') # typecasting from unsigned int to signed
bands_10.shape

This gives me a shape of (4, 10980,10980), which is teh shape of the total scene and not the subset

Answer 1

You can convert your coordinates to a row and col using the DatasetReader.index method, then build a window from that.

Below is a very simple example:

import rasterio 
from rasterio.windows import Window

# Check your coordinate ordering, 
# I wasn't sure which was X and which was Y
ulx, uly = 5773695.0, 601200.0  
width, height = 700, 500

with rasterio.open('/path/to/someraster') as ds:
    # Note - index takes coordinates in X,Y order 
    #        and returns row, col (Y,X order)
    row, col = ds.index(ulx, uly)
    window = Window(row, col, width, height)
    data = ds.read(window)