If I understand your question correctly, you want to train 2 neural networks simultaneously and then combine them into one through an element wise operation of the layers. You're not looking to merge two branches of the same network. Is my understanding correct? I'll reach out to you to learn more about your use-case.
By Looking at the workflow diagram, I am wondering if you're trying to use an encoder-decoder network for a semnatic segmentation task? If that's true, then learn more here --> https://www.mathworks.com/help/vision/examples/semantic-segmentation-using-deep-learning.html
If you're looking for combining two branches of a neural network using an element-wise operation, then we've 2 ways in which you can do that. 1. Depth Concatenation layer : https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.depthconcatenationlayer.html
2. Addition layer (as mentioned by Bernhard): https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.additionlayer.html?searchHighlight=addition%20layer&s_tid=doc_srchtitle