Well, I think the most important thing to be considered first is what features you would like to assess and compare about the two buildings.
From my understanding, certain architectural features do allow us to compare quantitatively, such as the quality of the building structure, workmanship of the finishes, degree of flexibility of the building layout etc. Even, the green passive design features of a building can be assessed and compared quantitatively. In order to enable the comparison to be done digitally, such features have to be quantified in the first place.
While for the architectural style, which I am most likely to refer to as aesthetic, is basically to be assessed qualitatively through perception survey.
The idea is rather new and could be an interesting issue to be studied through the fractal theory. Not an easy study, though, but it could be a good subject for a thesis or a new system presentation.
Chee raises what I think is the key question that needs to be resolved- what is it you are trying to assess? I think you need to clarify this in order resolve what I think you are asking in your question, " Could architectural styles of images of two different buildings be compared digitally?".
If you asking can individual features such as an ionic column be compared in two photographs, then using a process similar to face recognition, identifying the key data points and spatial relationship between them, then you could build up a database of 'architectural features'. It is the distinctive geometry of features, that makes them recognisable, I am thinking here of the ratio of height, width, etc.
If however, you are referring to the actual 'architectural style', such as Bauhaus, then this may be more problematical, depending on how much sub splitting you intend to do in terms of 'style'. There will be more crossover in individual features between styles such as Bauhaus and Vkhutemas, whereas it should be easier to identify between classical and modernism for instance.
If you are considering comparing quality of workmanship, identifying interventions that have occurred to the building, then I would anticipate this will be much more difficult. It is not necessarily the just 'skill' of the craftsman/builder that influences the appearance of the 'workmanship'. Other factors such as availability of quality comparative materials, budget, site constraints, weather, time allowed all impact on the 'quality' of workmanship. So to compare a modern repair carried out in 6 weeks to a building, which might have had the original area constructed over a number of years, is not necessarily comparing like with like. There are also ethical factors to consider such as was the repair intended to be seamless, or was it supposed to be an obvious repair, so that the feature can be 'read' as part of the history of the building.
I see that you have other questions regarding stylistic influences of architecture and how to interpret them. If you followed the methodology of identifying individual features columns, doors, windows, etc., for a variety of styles you are interested in, then you would be able to arrive at a quantitative rather than qualitative assessment of the number of individual features within your considered building. If it was applicable you could perhaps consider then identifying a boundary condition for a number of features that indicates little or no influence, say 0-3 features, small influence 4-10 features, etc.
Thank you very much Tim for a detailed answer. You are right that geometrical features, such as proportion, height, breath etc. are easy to recognize while the underneath style concepts could not be recognized. My question has been appropriately answered. Thanks again.