Hi
Your question is fairly vague so I have provided a fairly wordy answer.
You probably want to determine first, even as an outline, how you would measure the timings in a real browser via the javascript timing apis, boomerang or a commercial RUM soln. You can time the individual resources or create a custom user timing beacon for example.
http://www.w3.org/wiki/Web_Performance/Publications
As the user experience of this lazy loading could be largely affected by the browser and javascript performance, you should check that first also with tools like Web page test.
Per the docs, Gatling is not a browser. So you will need to include only the relevant details of the page and how it loads for perf test purposes in the gatling scripts.
You can group all the lazy images for reporting their response time, either individually, or the duration of all in each page for example.
http://gatling.io/docs/2.0.0-RC1/general/scenario.html?highlight=group#groups-definition
Similar to the other thread on getting timings closer to a real browser you should check whether the browser loads the images in parallel or sequentially and script accordingly ( modify the recorded script if needed).
The recorded script will not be dynamic. You may need to make it dynamic so it makes a representative range of image requests. If say you have 1M images over many pages on your site then requesting only 1 page's worth of images over and over may not be valid.
Thanks
Alex