In a computer based system we store images in 1D (linear array), which is interpreted as 2D. More over using stereo correspondence we will be able to extract 3D information aswell (stereograms). I will discuss the problems that we are going to face during the recognition process due to this kind of a storage. Assume that our brain stores information in 2D. In order to recognize an object from any direction, it will have to store a lot of 2D images. There are infinitely many objects that our brain can recognize, which means that it will have to store an innumerable number of images as well. Storing in 3D also means an innumerable amout of information to handle and process. More over to recognize an object how will we know how much to rotate it and which object needs to be rotated to match it? Now that is what is recognition! If you know which object needs to be rotated and by how much, what is the need to match? And if you don't know that, how do you proceed? Does our brain store information in any one of these forms? What about 1D. There is a lot of difference between a generalised object recognition and a specific one. When you are tailoring it for a specific one, you will know what you will be looking at, but once it is generalized, every pattern may be a valid pattern. Because you have to give room for your system to learn new things.