OpenCv Python UMat Image Stitching and Blending

Hello, I’m having some trouble performing Linear Blending using addWeighted on 2 images I am trying to stitch together in python openCV. The black space that is padded around the images to make them the same size, and to shift to allow the stitch, is causing parts of the image to darken
I managed to isolate the overlapping area via some matrix multiplication to create a binary matrix to generate only the overlapping area, but the process slows things down a lot. Also, I am using UMats to try and leverage the GPU, so my ability to interact with the with the matrices directly is limited to openCV functions, as far as I understand

  1. if I just throw addWeighted at the 2 images, i get this: