Get corners of images after performing rasterio merge

Question

I have a handful of georegistered raster images that I'd like to merge together with rasterio. All well and good:

mosaic, out_transform = merge(rasters_to_mosiac, method='first', nodata=0.0, resampling=Resampling.cubic)

However, I'd like to use graph cuts to stitch the images together along optimal seams using the GraphCutSeamFinder from OpenCV in Python. The method to estimate optimal seams requires knowledge of the corners of each sub-image. Are there are slick ways to achieve this that I'm missing?

Aside from any built-in methods, I think the following approach might work:

1. Resample all the input images to the same resolution
2. Get the bounds of each image, unary union them, and calculate the envelope geometry. This shape is the bounds of the final mosaic.
3. Calculate the expected width and height of the mosaic image based on the input resolutions and the calculated bounds.
4. For each image, normalize the top-left corner coordinates to the mosaic image bounds and calculate the image coordinates of the corner.