The performance in Chrome is noticeably better than Firefox. The smaller the window size, the smoother the playback. Since it is only four images at once, whenever one image leaves the scene and a new one appears, there might be a small lag when the image is loaded into memory. I don't have a deeper understanding on how this could be optimized further for the browser.
For the Android OpenGL implementation I was able to take care of loading the images asynchronously before they enter the scene and freeing up the memory again after they leave the scene.
Happy to take suggestions on how this might be optimized further, as it would be interesting to work with even higher resolutions for future projects, but this quickly leads to much higher performance load and choppy playback on weaker machines.