Image combination is possible; it depends on what you are looking for after conbination.
for example when you use the XOR operator to combine two images;the output image can be used to measure the similarity degree or the difference degree of input images.
You can use the technique of image blending. Both the images should have some common points of interest and you can create image pyramids and then blend all pyramids of both the images. I can share matlab code if you want
Actually averaging is preferred to avoid saturation since most images have 8-16 bit dynamic range. But usually wavelet or laplacian based techniques are preferred not to lose contrast at edges.
Some methods used for accelerating the MRI data acquisition, are based on the combination of some pixels of two or more images in the frequency space (Fourier domain) (XX% of information each image), and the next step is apply the inverse fourier transform, for generate the new combined image with the information in frequency of two or more images. Probably this idea can be useful.