I discovered today the following article, which mentions my API. An interesting approach they propose is, if I understand correctly, to use some common .NET class to load the HTML, and somehow extract the images from the HTML. When I wrote the API, some 2-3 years ago, I searched for such a thing but didn't find any - maybe I missed it?
If this works, it can remove completely the API's major liability, which is the dependence on the regular expressions. Right now, the API parses the HTML response returned by Google and when the format of this response is changed - the whole thing breaks. On average, since I initially published the API, the response format changed 1-2 times a year.The downside, of course, is performance - loading the whole HTML will always be much more CPU and memory intensive than using a regex. Yet, for most applications I guess it's a price that can be paid. Once I have a few spare hours I'll check it out. Or maybe next time Google change the response format and I need to dig into it again. We'll see.