We aim to build an automatic face identification system which is robust against the image variations caused by changes in head orientation, facial expression, lighting conditions etc., We previously developed an automatic face identification system based on a pattern matching method that uses 2D intensity images. The system is composed of three major parts : (i) automatic facial feature extraction which normalizs face size and position in the target image, (ii) extraction features for classification based on Fourier transform and KL expansion, and (iii) classification by Minimum Distance Decision rule. This paper reports the experimental system for automatic front-view face identification. The experimental system is able to automatically identify, within restricted head orientation, lighting and facial expression conditions, several people with high accuracy.

View full abstract

Last year, CCITT announced the recommendation for "Video Codec for Audiovisual Services at p x 64kbit/s", called CCITT Rec. H.261. It prescribes a transmission image which is called CIF. Standardized TV-telephone equipment must transform original TV images to CIF images. A conventional method to get CIF images uses scale reduction filters. However, this method requires amount of hardware. Another method is cutting original images with CIF rectangular outline. This method is difficult to detect and track a human head area. This paper proposes a new head area tracking method in order to get CIF images. A human area will be tracked robustly another moving object invades to original images. This paper describes a basic theory and first simulation results.

View full abstract

The distortion of the mandibular movement path is a characteristic symptom in mandibular dysfunction. On examinations and diagnoses of the mandibular dysfunctions. accurate measurement of the mandibular movement path is one of the available method. This paper described a method of measurement of the mandibular movement path. A subject is put one white mark on the upper tooth and the other on the lower tooth. And the mandibular movement of the subject is memorized as a moving image data. The mandibular movement path is computed from the data. And the simulation results show that the mandibular movement path is obtained.

View full abstract

On orthodontics or orthopedics. a treatment of a patient, who is the bone deformity of jaws, is done using the visual methods. For example, an esthetic line which is used as evaluated value for projection ratio of lips, a picture of profile and so forth. These visual methods are used to plan the treatment of a patient. We investigated the computer system aided the treatment of patient with an image processing. First, an esthetic line is obtained from an image data of profile. Second, a treatment effect is simulated with an image data of profile using a spline curve. This system were build in a personal computer.

View full abstract

In this paper, we propose a method of 3-D shape measurement using shading images of an object with unknown reflectance. Two cameras are disposed symmetrically with respect to the direction of incident light on the object surface. Values of brightness of the surface, accepted by two cameras, are equal when the azimuth direction of surface normal is parallel with the direction of incident light. Utilizing this property, we determine the azimuth angie of surface normal rotating the object, The object surface is reconstructed based on the azimuth angle. A measurement result of an uneven object is shown. Errors are less than 3 mm when the object is about 170 cm distant from cameras.

View full abstract

We devise algorithms to remove whitishness involved in the mesh-printed matter by a drum scanner or CCD image scanner. In the case of the single color image with a black background, we choose the pixel of minimum value among the four pixels, replacing the three with it. In the same case mentioned above and the position of the color pixel region being already-known, we obtain minimum value in the region and replace all the pixels with the minimum one. In the case of the girl with a hairband, we take the difference in absolute value between the scanned pixel and the next one, and we replace the latter with the former if the absolute value is between a certain value. By these three operations, we can get satisfactory results in removing whitishness.

View full abstract

We analyze frame rates in low bit rate video coding and show that an optimal frame rate can be theoretically obtained. In order to achieve an optimum balance between coded picture quality and motion smoothness, a frame rate has to be selected which is appropriate to parameters such as coding scheme, property of the video signals and coding bit rate. In this paper, coding distortion measured by mean square error due to quantization is analyzed by modeling video signals, and it is shown that the frame rate can be uniquely optimized. The result of this analysis is compared with the results of computer simulation. In addition, the relation between this analysis and the subjective evaluation is described.

View full abstract

In a DPCM system, when a buffer reglstor ovarflows, a quantization table is sometimes switched from fine to coarse. Therefore, the degradation of reconstructed images will occur. In this paper, the method to improve the dagradation of reconstructed images is presented. The method is simple, that is, reconstructed images after the quantization table is changed are replaced by the average of a present image and the provious image. For motion plotures, the motion componsation technique is effectively adopted.

View full abstract

All-in-one type videophones are very much in demand. In order to supply the demand, a compact H.261 video codec is indispensable. An architecture named hybrid archltecture is proposed. The hybrid archltecture which consists of hardware and software is sultable for a low bit rate codec in point of reducing the codec size. Based on the archltecture, a A4 size video codec is developed. The developed codec is also illustrated in this paper. For the hybrid archltecture, evaluation schemes of the coding and decoding abllity are proposed. With the schemes the ability of the codec is estimated.

View full abstract

A development of video storage and transmission equipment is presented. First, requirements for this equipment are discussed from a standpoint of video handling view such as displaying, editing. Next, the techniques to achieve these requirements are described. Then specifications of developed equipment are described. This equipment consists of video/audio codecs, digital storage media (magnetic hard disc) and a terminal adapter for N-ISDN. It is controlled by the workstation and has three major functions ; normal and semi reverse playback and limited random access, editing of stored video signals, high-quality video transmission through a low speed channel by using longer transmission time than coding time.

View full abstract

A reduction of the number of transmitted frames can often be adopted in a low bit-rate coding (64kbit/s) of videophone images. However, the transmitted images may have a lot of jerkiness, if the period between the transmitted frames is long. This paper, from the viewpoint of application of model-based image coding techniques to waveform image coding, proposes a new frame interpolation method for moving images based on a 3-D-shape model of an object and its 3-D motion vector. Firstly, the basic framework of the proposed frame interpolation method is presented. Secondly, the 3-D motion estimation method and the frame interpolation method using a 3-D model are described in detail. Finally, we give experimental results showing that the proposed method is an effective frame interpolation method.

View full abstract

For transmitting and receiving high quality color image with 64 kbps which is available at stabilization of power systems and contingent troubles occurring in power supply facilities. We have developed the color bubble jet facsimile terminal aimed at high resolution color image transmission. It can transmit a binary color image data of maximum image size A3 at 400 dots/inch with 4 bits assigned to each picture element and it has monocolor G4 facsimile mode and copying capability. It is capable of transmitting an A4 document in less than 30 second.

View full abstract