get screen width and height in world size

Hm, if you just want to check if the object is visible to the camera, then you can use:

MonoBehaviour.OnBecameVisible() http://docs.unity3d.com/Documentation/ScriptReference/MonoBehaviour.OnBecameVisible.html

MonoBehaviour.OnBecameInvisible() http://docs.unity3d.com/Documentation/ScriptReference/MonoBehaviour.OnBecameInvisible.html

You can also use:

GeometryUtility.TestPlanesAABB(...) http://docs.unity3d.com/Documentation/ScriptReference/GeometryUtility.TestPlanesAABB.html

with: GeometryUtility.CalculateFrustumPlanes http://docs.unity3d.com/Documentation/ScriptReference/GeometryUtility.CalculateFrustumPlanes.html

Then you can test the objects bounding box against the camera frustrum.

 using UnityEngine;
 using System.Collections;
 
 /// <summary>
 /// Update camera near plane world coords.
 /// This could be repurposed to handle camera occlusion 
 /// Directing some event (such as moving the camera) to occur 
 /// rather than the debug messages.
 /// </summary>
 public class UpdateCameraNearPlaneWorldCoords : MonoBehaviour {
 
     // properties
     public Camera theCamera;
     public GameObject player;
 
     Vector3 checkpoint; // player transfrom that we offset from 
     Bounds bounds; // player bounds for positioning sensible check point.
     Vector3    p; // point in space
     Vector3 trajectoryEndPoint; // where we want to check we can see in world space
     Vector3 trajectoryStartPoint; // The "checkpoint" with a useful offset.
     Vector3 trajectory; // the direction of the ray to fire.  ( I invert it here to be fair)
 
     static RaycastHit hit;  // Stop it needing to be declared a bazzillion times.
 
     // Helper Method
     static bool CheckIsOccluded (Vector3 trajectoryStartPoint, Vector3 trajectory, Vector3 trajectoryEndPoint)
     {
         // Simply return true or false.  True means we are occluded.
         // We could have used...
         //return Physics.Raycast (trajectoryStartPoint, -trajectory, out hit, Vector3.Distance (trajectoryStartPoint, trajectoryEndPoint));

         //but this makes more sense in this instance
         return Physics.Linecast(trajectoryStartPoint, trajectoryEndPoint, out hit);
     }
 
     // Use this for initialization if you are networked or something
     void Start () {
 
         // Could find objects dynamically is we are running NetworkView stuff. via isMine.
         // In this example The PLAYER and CAMERA are simply dragged on to get the reference
 
         // You might have a mesh filter or w/e - get the right one from the right level (Self/Child).
         bounds = player.transform.gameObject.GetComponentInChildren<SkinnedMeshRenderer>().sharedMesh.bounds;
 
 
     }
 
     void SetupValuesReadyForCornerCheck (float x, float y)
     {
         p = theCamera.ScreenToWorldPoint (new Vector3 (x, y, theCamera.nearClipPlane));
         trajectoryEndPoint = p;
         trajectory = trajectoryStartPoint - trajectoryEndPoint;
         Debug.DrawLine (trajectoryStartPoint, trajectoryEndPoint, Color.red);
     }
     
     void Update () {
 
         checkpoint = player.transform.position;
 
         // Make this be somewhere useful to you. This place is about right for me.
         trajectoryStartPoint = (checkpoint - bounds.center) + (Vector3.up *0.4f) ;
 
 
 // TOP LEFT
 
         SetupValuesReadyForCornerCheck (0, Screen.height);
 
         if (CheckIsOccluded (trajectoryStartPoint, trajectory, trajectoryEndPoint)) {
             //Vector3 targetPosition = hit.point;
             Debug.Log("Occluded Top Left");
         }
 
 // BOTTOM LEFT
 
         SetupValuesReadyForCornerCheck (0, 0);
 
         // Check for hit.
 
         if (CheckIsOccluded (trajectoryStartPoint, trajectory, trajectoryEndPoint)) {
                         //Vector3 targetPosition = hit.point;
                         Debug.Log ("Occluded Bottom Left");
                 }
 
 // TOP RIGHT
 
         SetupValuesReadyForCornerCheck (Screen.width, Screen.height);
 
         // Check for hit.
         
         if (CheckIsOccluded (trajectoryStartPoint, trajectory, trajectoryEndPoint)) {
             Debug.Log("Occluded Top Right");
         }
 
 // BOTTOM RIGHT
 
         SetupValuesReadyForCornerCheck (Screen.width,0);
 
         // Check for hit.
         
         if (CheckIsOccluded (trajectoryStartPoint, trajectory, trajectoryEndPoint)) {
             //Vector3 targetPosition = hit.point;
             Debug.Log("Occluded Bottom Right");
         }
 
 
     }
 }
 

Is not far off the mark. Could doubtless be more efficient.

EDIT: if you want to see what is happening drop in a

 Debug.DrawLine(trajectoryStartPoint, trajectoryEndPoint, Color.red);

Inside the helper occlusion check. (pictured, with some extra GUI texts for clarity)