Kazuhito Murakami

On KANSEI facial image processing for computerized facial caricaturing system PICASSO

The PICASSO system for facial caricaturing is a typical KANSEI vision system in the following senses: it can process 2D, 3D and motion facial images to generate a facial sketch of line drawings; it can extract individuality features by introducing a mean face for enforcing the visual KANSEI impressions by which the face looks more likely to be so; and it can evaluate the generated facial caricature to be moderate for the individual visual KANSEI characteristics, based on the model of visual illusion. Image processing methods for generating a facial sketch based on the Hough transform are presented. Principles to generate a facial caricature and to evaluate the generated works are explained precisely, together with demonstrations of experimental results. In order to equip the PICASSO system for the feedback loop from the observer, we have introduced the eye camera system. This equipment is expected to provide a new method to extract the simultaneous individual eye-mark patterns which show where in the face the observer is watching. These preliminary experiments are also presented to show the possibility to extract more tightly individual KANSEI visual information.

Facial caricaturing with motion caricaturing in PICASSO system

The paper discusses how motion features of emotion expression processes could enforce facial caricaturing. It is known from experiments that there are various kinds of motion features in the laughing processes. We implemented a system to deform the motion features caused by emotional expression for enforcing facial caricaturing. In this paper, we propose a new idea to extend the extraction mechanism of the individuality features to the time-axis in addition to spatial extension.

From coarse to fine correspondence of 3-D facial images and its application to 3-D facial caricaturing

We propose a method to generate a facial caricature depending on a simple method for corresponding 3-D facial images. In this method, we extract several facial parts regions each of which include facial parts of the face by using both gray and range images. The diagonal corners of the respective extracted regions are used to provide the informations on the correspondence between the faces. Therefore, by using this method, it is expectable to reduce the number of the feature points for the correspondence from more than several hundreds to ten or so. At the same time, this method makes it possible to extract boundaries of facial parts. In order to examine the feasibility of this method, we employ a usual corresponding method of the triangular patch as the reference. We generated the 3-D mean face and 3-D facial caricature to demonstrate experimentally the feasibility of the proposed method. It was clarified that the number of the correspondence points can be reduced to only 10% of the usual method.

Facial caricaturing system PICASSO with emotional motion deformation

Discusses how motion features of emotional expression processes could enforce facial caricaturing. A still caricature is not enough to express individual character. It was experimentally known that there are various kinds of motion features in the laughing processes. We implemented a system to deform the motion features caused by emotional expression for enforcing facial caricaturing. We propose an idea to extend the extraction mechanism of the individuality features to the time-axis by means of the in-betweening method. It is clarified that individual facial features can be enforced by emotional motion deformation in caricaturing.

Global feature extraction using efficient Hough transform

Novel Hough transform schemes are presented to provide two kinds of function expected of a global feature extractor. While the Hough transform was originally no more than a line segment detector, more complicated pattern features can be extracted by more intensive investigation of the Hough parameter space and/or more hierarchical usage of the Hough transform. As a typical example, a method to extract convex hulls and other related features from the transform space is introduced. In order to reduce the computing cost, a few efficient algorithms are introduced. A family of the functions applicable to the Houghlike feature extractor, the extended Hough transform, is proposed. As an example of hierarchical usage, the Hough transform is applied to the feature points of the Hough space.<<ETX>>

Face recognition through hough transform for irises extraction and projection procedures for parts localization for facial caricaturing system PICASSO

In order to generate facial caricatures automatically, it is necessary to recognize facial parts in advance. Since it is difficult to realize the face recognition, the primal description of face has been provided manually so far also in our project PICASSO for facial caricaturing.

Re-generation of facial image eye-contacting with partner on TV conference environment

We have been developing a method for recognition of facial parts such as eyes, nose, mouth, etc. The paper concentrates on eye (iris) recognition using a Hough transform. Based on this iris recognition, the paper proposes a basic idea and method for re-generating a facial image which is making eye-contact with a partner in a TV conference environment. We propose a new methodology for realizing a natural human interface of face-to-face communication in a TV conference environment. Since a human is likely to look at the face of her partner on a TV monitor and not at the TV camera, she will usually fail to send her own eye-contacted facial images. The basic idea of the paper is to solve this unfortunate communication degradation by regenerating a facial image by changing the direction of the irises in the same original facial image. The proposed methods and experimental results are presented to demonstrate the effectiveness. Future problems include improving the regeneration method of irises, introducing a real time processing method, coping with the motion of the head, and finally proposing a new eye-camera system which could reduce the constraints caused by head mounted goggles. This new method of eye-camera is also introduced.

On Typical Implementations of Hough Transform for Improving Its Performances

This paper proposes two typical implementations of Hough transform for improving Hough performances. RVHT(Randomized Voting Hough Transform) can provide higher detectability of shorter edge lines with lower computation cost than RHT(randomized HT) and PHT(probabilistic HT) algorithms.

An Autonomous Facial Caricaturing Based on a Model of Visual Illusion - Experimental Modeling of Visual Illusion

In the PICASSO system, a facial caricaturing system, some visual illusions such as the Wundt-Fick illusion and the Ponzo illusion for example, are applied to evaluate the shapes of the facial parts such as eyebrows, nose, mouth and face contour, in the deformation process of caricature generation. In many cases, as well-deformed caricatures are evaluated to be successful, it is confirmed that the utilization of the visual illusion is effective to evaluate the results of caricatures. In this paper, the models of mathematical expressions of visual illusions are presented together with the results of human experiments. And, a method to apply visual illusions and to control the exaggeration process is proposed.

On autonomous control method for facial caricature generation based on a model of visual illusion-experimental modeling of visual illusion

In the PICASSO system, a facial caricaturing system, some visual illusions are applied to evaluate the shapes of the facial parts in the deformation process of caricature generation. In many cases, as well-deformed caricatures are evaluated to be successful, it is confirmed that the utilization of the visual illusion is effective to evaluate the results of caricatures. In the paper, the models of mathematical expressions of visual illusions are presented together with the results of human experiments. A method to apply visual illusions and to control the exaggeration process is proposed.