Questo sito utilizza cookie di terze parti per inviarti pubblicità in linea con le tue preferenze. Se vuoi saperne di più clicca QUI 
Chiudendo questo banner, scorrendo questa pagina, cliccando su un link o proseguendo la navigazione in altra maniera, acconsenti all'uso dei cookie. OK

An Exploration of the Introduction of Warping to the Eigenfaces

The aim of this work is to provide a robust and reliable model of the face by exploring the benefits of the introduction of warping as a pre-processing technique to improve the result of the eigen faces statistical analysis. We prove the advantages of such an approach both from a theoretical and experimental perspective.
This model is based on all the previous works listed but it has a different approach to the problem of modelling the face space : on the theoretical side we try to eradicate the sources of noise and variance before performing the face analysis rather than trying to minimise them during the process. From the practical point of view
we introduce a complete modularity of the processes therefore we reduce the risk of side effects and correlation that might reduce the reliability of the results. Moreover we can add other pre-processing or post processing techniques to improve the results without affecting what we have already achieved.
The key issue is the eigen face as everything we studied and all this work has been done to improve the performance of this technique but we prove that this technique is a valid and reliable approach to the problem.
We approached both the shape variance and the lighting variance from a slightly different perspective to create a technique that is widely usable and can be adapted to any particular set of data by changing one of the two algorithms involved leaving the other one unaffected.
We have applied the "Douglas Smythe" ( 1990 ) algorithm to warp a training set of 21 colour images as a pre-processing technique of the eigen-faces analysis.
This approach is motivated by the intrinsic nature of the eigen-faces where all the important features of the faces are supposed to be aligned. In real life this is quite unlikely to happen and the result is blurred. But, if we apply a warping technique to all the images of our training set and we warp them to any arbitrary image, within this set, before performing the eigen faces; the images will have exactly the same features in the same spatial coordinates. This should dramatically improve the performance of the eigen technique.
Moreover we have a set of landmarks and offsets that can be used to perform a statistical analysis to describe how these point vary in our training set that describes the spatial variance of the main features of the faces in the training set we use. The applications are various: we can use this program as a platform to achieve image recognition or to create a large number of new faces from the training set or we can use it to add and remove facial expressions, locate a face in an image and to achieve a more "natural" animation

Mostra/Nascondi contenuto.
1. INTRODUCTION The face has always played an important role in any society. Every human can easily detect, recognise and memorise thousands of different faces. This skill is robust enough to allow us to perform those tasks without being affected by radical changes in the viewing condition or variations in the gender, age, race or geometric orientation of the face. We can also perform accurate analysis and find analogies in those faces without a big effort to guess the age of a face or its expression. It would be remarkably useful if we could perform some of those tasks using a computer system but, unfortunately, the face is quite a complex object to model and to work with. There are two different approaches to the problem: a natural one and a synthetic one The natural approach uses a certain number of faces images to find a certain amount of variations through the face space and will create some sort of prototype face based on the physical appearance of the faces. This is usually obtained by performing some statistical analysis. The synthetic approach tries to model the face by deducing some mathematical functions that can approximate it. Usually, as it s almost impossible to find such a global function, the task is performed by splitting the face space in a collection of simpler spaces such as the nose, the cheeks, the eyes etc. and trying to find suitable functions to describe these easier problems and to synthesise those features. I have called it synthetic because it usually loses the richness of the grey level information associated with a real image in order to find a codable function. These approaches can be naturally used in a 2-D space and in a 3-D one. The 3-D approach is more natural as the face is located in a 3-D space but it is a complex task to model a face object in this space. There are different methods to cope with this complexity but they have a high computational cost to achieve a result that usually is not robust. They tend to use the synthetic approach as the cost of a complex statistical analysis in a 3-D world is not computationally feasible. The 2-D approach might look like a reduction of the problem but it is more attractive for its reduced computational cost and the result we obtain is not affected by this flattening in a 2-D space as the analysis in Psychophysics shows that our perception of the face is projected in a 2-D space in our mind. The modelling can also be obtained more straight forward making the development of a successful technique much more intuitive.

Tesi di Laurea

Facoltà: Scienze Matematiche, Fisiche e Naturali

Autore: Luca Sorbello Contatta »

Composta da 54 pagine.

 

Questa tesi ha raggiunto 496 click dal 20/03/2004.

Disponibile in PDF, la consultazione è esclusivamente in formato digitale.