I think that the easiest and more efficient approach is using normals or facing ratio. But I tried to recreate something more physically correct. Don't get me wrong, I'm not saying the other is dead easy, I'm sure it's tricky, but I thought recreating the Lenticular Lens would be more challenging.
First of all, I did some research and found this great website https://www.lenstarlenticular.com/the-basics-of-lenticular you can find there all the information needed for this Challenge. As we are rendering 2D images, I focused on the animation effect rather than the 3D effect (if someone has VR set, would be nice to try it too). Animation effect uses a wider lens while 3D effects uses narrower.
I started doing a RGB grid to the use as a mask and a different image per channel.
Then, a 45 degrees angle gradient (as the website suggest for Animation effect), and also made it a grid.
This will be used as a displacement map on the top face of the geometry, while the images are going to be displayed on the other face.
Will be crucial to play with the displacement height. And be sure that the gradient map is perfectly matched with the grid of 3 images.
The whole object will be refractive except from the back face, it will be a basic shader with grid images.
The scene still needs some polishing and more render passes, there's some flickering.
Here you have the result, 3ds max + Arnold