Before making any recommendations you should review SharePoint 2010 best practices for Search erformance. Articles like this one (even though they were written in the beta timeframe) hold valuable insight into large list/library handling. http://blogs.msdn.com/b/enterprisesearch/archive/2010/06/09/sharepoint-2010-search-dogfood-part-3-query-performance-optimization.aspx
Using Query is always going to be faster that any other approach. The only issue is "freshness" of the index. This means ensuring that you are configuring search to crawl the corpus to keep the index fresh. Whether you chose query (and which query syntax you choose) over another approach will also be determined by how you are creating your search experience for your end users.
Note that in SharePoint 2013 the Search.asmx and SQL Query Language approaches are depricated in favor of REST and Keyword Query Language. So if you are planning on implementing a SharePoint 2013 Search farm anytime soon, that may drive your decision. Note that you can implement SharePoint 2013 Search against 2010 farms and gain the advantages of Continuous Crawling and significantly improved search experience without having to upgrade your 2010 site collections.
BTW, 100,000 is not that "huge" particularly for images, which tend to be slim on metadata. You generally plan additional index partitions at 10M documents. Are you planning to employ a custom iFilter to handle the image metadata or just go with the out of the box null filter?