Acquisition and rendering of Lightfields

Description:

Among the various techniques of Image Based Rendering, Lightfield Rendering has been of special interest. The primary reason behind this is that no depth information or feature matching is required to generate novel views. The process of rendering just involves combining and resampling the given images. Acquistion of image data however remains a challenging and time consuming task. In most methods expensive, complex and bulky setups are being used.

Here we describe a simple method of acquiring the image data set. The method requires a normal handheld video camera, which is taken around the object to be rendered. The simple computations involved make the method suitable for online lightfield acquisition.


Reports

  1. An Uncalibrated Lightfield Acquisition System
    Puneet Sharma, Angshuman Parashar, Subhashis Banerjee and Prem Kalra.
    ICVGIP 2002.
  2. Extended version submitted to Journal of Image and Vision Computing

Setup

Our experimental setup involves taking a handheld video camera around the object for which the lightfield is to be created. The object is placed between two planes, the frontal plane is transparent and both the planes have four identifiable points each, marked on them. The setup is shown below.


Rendering results

Nearest Neighbour Approximation vs Quadrilinear Interpolation

Nearest Neighbour Approximation Quadrilinear Interpolation
Nearest Neighbour Approximation Quadrilinear Interpolation
Nearest Neighbour Approximation Quadrilinear Interpolation
Nearest Neighbour Approximation Quadrilinear Interpolation
Nearest Neighbour Approximation Quadrilinear Interpolation

Clay Dog Model



Lord Ganesha



28th June, 2002