Accessing camera "backbuffer" in surface shader

Hi Everyone - this is my first question here, so I hope I'm following all the rules. Checked the existings, but couldn't find anything, so I'm creating my own.

Essentially, I'm trying to execute a refraction shader that utilizes output from a camera (based on this gpu gem: http://http.developer.nvidia.com/GPUGems2/gpugems2_chapter19.html)

What is the best (most performant/efficient) way to pass camera render data to a surface shader as a texture?

I've got a really rough experimental start using C#, with the following script attached to a camera:

 using UnityEngine;
 using System.Collections;
 
 public class camPostRenderScript : MonoBehaviour {
     
     public Renderer display;
     Texture2D tex;
     
     void Start () {
         tex = new Texture2D(Screen.width,Screen.height);
     }

     void OnPostRender() {
         tex.ReadPixels(new Rect(0,0,Screen.width,Screen.height),0,0);
         tex.Apply(false);
         display.material.mainTexture = tex;
     }
 }

However, this has drawbacks in that it only references one screen object at a time, and using C# instead of shader code feels like it's taking work off of the gpu and putting it onto the gpu which I think has performance implications (if any experts want to speak on that, I'm all ears).

Is there a way to create a shared texture that all of the refraction shaders will be able to access? I suppose I could make that "tex" var accessible, but is there a way to pull that in with a surface shader? Is it super heavy to write per frame like this? I'd imagine it's cheaper than using RenderToTexture() - is that correct? I'd prefer to avoid using RenderToTexture as I get the impression that it's pretty heavy. Is this method faster/lighter?

Alternately, is there another, "shader native" way to access a camera "backbuffer" like this that might be more efficient/live on the gpu?

Any input would be greatly appreciated :)

Thanks Everyone!