No abstract

This paper proposes a method which analyzes and synthesizes image sequences. and updates a three-dimensional facial model in knowledge-based image coding. Rules for synthesizing the output images are formulated to simulate the facial muscular actions. The input image analysis technique estimates the head motion and the facial actions direclly and robustly without any correspondences. This technique also provides good reproduction of the original images because it is incorporated with the synthesis rules. The presented method for updating the facial model improves analysis accuracy and quality of the synthesized images.

A technique for realizing linear-phase infinite impulse response (IIR) filters has been proposed by Powell and Chau and gives a real-time implementation of ( 1 ) ( ), where ( ) is a causal IIR filter function. In their system, the input signal is divided into -sample sections, time-reversed, section convolved with ( ), and time-reversed again. The signal is then filtered by ( ) to give the system output with a processing delay of 3 + 1 samples. However, the group delay response of the system exhibits a minor sinusoidal variation superimposed on some constant value. This variation will degrade image quality in image processing and signal quality in signal transmission applications. Furthermore, the output of the system contains harmonic distortion for a sinusoidal input. The main drawbacks of Powell and Chau's technique are the large processing delay of 3 + 1 samples and the accompanying phase and harmonic distortions. A smaller processing delay increases the phase and harmonic distortions, yet the amplitude response remains acceptable. Previously, the present authors presented a method of reducing the processing delay by shortening the section length by an integer factor using a structure with increased number of paths for the time-reversed signal. In this paper, the authors consider how to reduce the phase and harmonic distortions. We examined the operation of the sectioned convolution and analyzed it based on a state-space representation. Then, we found that the cause of the distortions is a periodic variation of the impulse response length in the sectioned convolution. To overcome this problem, a technique is devised to realize a recursive circuit having a truncated impulse response with a fixed-length . A system applying this technique to the Powell-Chau system is demonstrated to exhibit perfect linear-phase characteristic and produce virtually no harmonic distortion. Therefore, the section length can be reduced without limitation due to phase and harmonic distortions. Two methods for reducing the increased computational complexity of this technique assuming fixed are developed, and simulations are performed for the proposed system to confirm the expected improvements.

SUMMARYThis paper proposes a method which analyzes an expression from a facial image using a three-dimensional facial model and then extracts the facial expression.First, the head motion and the facial actions (such as those of eyebrows, eyes, and lips) are separated from the facial image, This is realized by estimating the threedimensional motion of the face based on the threedimensional facial model and by c o w s a t i n g the motion.Next, the expression information is extracted from the separated facial actions in two ways. One is the method to extract successively the facial expressions considering the characteristics of the facial actions based on the facial muscles. The other is the method to estimate the facial expression as a whole using the leastsquare method and regarding the facial actions by the facial muscles as a vector. Thorn methods are combined with the expression synthesis rules. This makes it possible to reconstruct the original expression from the extracted facial expression parameters.Finally, the result of the analysis of the facial expression from the actual image is compared to the result of evaluation by a psychologist to demonstrate the usefulness of the proposed method. The image reconstructed from the result of analysis also is compared with the original image.

No abstract