Files containing 3D objects, typically represented at 3D meshes with certain geometry and texture information, are very large. Therefore, not only do 3D objects take a lot of storage space, it is also extremely time-consuming to transmit them over the network for visualization. In addition, most 3D visualization applications need the entire 3D data file to render the 3D object even though the user may be interested in only a small part or a low-resolution version of the object. Progressive coding of 3D objects can resolve these problems. In this paper, we report our recent progress in progressive representation, transmission, and visualization of 3D objects. In our scheme, both geometry and the texture of the 3D object are progressively coded and transmitted. More perceptually important information is transmitted before the lens important information, which allows the user to stop the transmission at any time and yet retain the best available perceptual quality of the object at that time. Furthermore, the visible portion of the object is transmitted first and the non-visible portion is transmitted later, or not transmitted at all, in order to save the overall bandwidth.
This paper gives an overview of the research and development activities in 3D surface digitizing and modeling conducted at the Industrial Technology Research Institute (ITRI) of Taiwan in the past decade. As a major technology and consulting service provider of the area, ITRI has developed 3D laser scanning digitizers ranging from low-cost compacts, industrial CAD/CAM digitizing, to large human body scanner, with in-house 3D surface modeling software to provide total solution in reverse engineering that requires processing capabilities of large number of 3D data. Based on both hardware and software technologies in scanning, merging, registration, surface fitting, reconstruction, and compression, ITRI is now exploring innovative methodologies that provide higher performances, including hardware-based correlation algorithms with advanced camera designs, animation surface model reconstruction, and optical tracking for motion capture. It is expected that the need for easy and fast high-quality 3D information in the near future will grow exponentially, at the same amazing rate as the internet and the human desire for realistic and natural images.
3D reconstruction technique plays an important role in the applications for 3D data acquisition, such as medical diagnosis, animation and virtual reality. Moreover, the 3D triangulation process is one of the most important parts while reconstructing the smooth surface of a 3D object. The essential of 3D triangulation is to find the intersection of the rays emitting from the correlated points on each image pair, but the emitting rays always don't intersect with each other owing to the error in process. So the obtained 3D point is an approximated value and makes the reconstructed surface uneven. In the paper, we proposed an triangulation enhancement method, which reduces the perturbation in the reconstructed data and filter out the error caused by spurious vectors in the process of correlation.
Facial model generation is an important issue in the model- based applications, such as MPEG-4 and the virtual reality. An effective and precise construction algorithm of 3D facial model from 2D images should be available for practical applications. To generate facial model usually requires stereoscopic view of the face in the pre-processing stage. Although facial model can be successfully estimated from two stereo facial images, the occlusion effect and imprecise location of the feature point prohibit us from obtaining an accurate facial model. In this paper, we proposed several facial model estimation algorithms to find the precise facial model form a stereo or mono image sequence. The information of head movement, which is recorded in the image sequence, in the temporal domain is utilized for the facial model estimation. Even though the a prior information about the 3D position of the head with respect to the camera and the rotation axis and angle of the head's movement are unknown, an accurate facial model (within 7.21%) error) can still be obtained by our schemes. In addition, our schemes do not require the precise camera parameters and avoid the tedious camera calibration such that the facial model generation is easily achieved.
Vision is an ideal sensor modality for intelligent robots. It provides rich information on the environment as required for recognizing objects and understanding situation in real time. Moreover, vision-guided robots may be intelligent and largely calibration-free, which is a great practical advantage. Together with it, a new concept for intelligent robot control, that enables realization of the calibration- free visual robots, is introduced.
This paper describes an automatic face recognition algorithm for security entrances. There are two major steps in this procedure to make the automatic recognition possible: (1) We combined the two-phase face detection method and back propagation neural networks to detect human faces when people are walking in the region of entrances. The combination allows the strength of both methods activated to accommodate the size and head-orientation variations and to eliminate the false detection. (2) Novel face recognition: we extract the facial feature measurements to form the multi-variable normal distribution for each person. These multi-variable normal distributions separate the decision space well and the probability for good index for face recognition. This face recognition algorithm is very efficient on computing time and taking little storage space.
This paper presents the design and performance of Chung Shan Institute of Science and Technology newest man-portable infrared imaging system, the Seagle-1. The thermal imager is designed for day and night long-range observation and forward reconnaissance and surveillance applications. The camera system achieves an NETD equals 0.067 K at 30 Hz frame rate with f/1.8 optics (300 K background). The design and performance of the 256 X 244 PtSi infrared camera will be described in this paper.
This paper explored the design of large ink supply system of ink-jet printer, which is in general installed in so-called wide format printer today. Subsequently, a new type of large ink supply system was presented to fulfill the fundamental functions that ink in the reservoir could be automatically delivered into the print head of printer by means of capillary force in nature. Moreover, the new system was characterized with no traditional long tubing portions such that pressure loss or vibration, due to long ink passage of tubing, could be eliminated. To achieve the goal of removing traditional long tubing system, the ink reservoir width of system must be greater than the print width of printer. As a result, a stable back pressure of print head can be kept all the time no matter the print is printing or not; it could also be more stable than before even if print head is moving in high speed. Hence, better print quality could be obtained in the printer equipped with the new system of paper.
Meniscus oscillation usually occurs after jetting drops in Drop-on-Demand inkjet head. It is important for ink refill motion. Ink refill motion affects many jetting performance in inkjet technology. Therefore it is very important to understand ink refill motion process for designing inkjet head dimensions and ink properties, such as nozzle diameter, barrier thickness, ink viscosity and surface tension. This meniscus oscillation is a kind of under-damped oscillation. The refill time is defined, as the time required returning to the initialization, the free surface obeys damped oscillation and oscillates between meniscus mounding and recession when the amplitude of the oscillation reduces to dispersion. This study researches several inkjet heads using computational fluid dynamics simulation. CFD solver, such as FLOW_3D, is used to solve the problem. After calculation, the results are plotted with post-processor such as plot software and output to printers. This paper shows the break- off time, the refill time and operation frequency of the inkjet head.
3D image display devices have wide applications in medical and entertainment areas. Binocular (stereoscopic) imaging without glasses, especially spatial-multiplexed displays such as lenticular display, barrier strip display, and single-lens stereoscopic display, is one of the most powerful and popular ways for life-like presentation of our 3D environments. The definition and relationship of the image contrast and viewer crosstalk are reviewed and clarified. They are measured and compared on three different types of 3D display systems, including shutter-glasses stereoscopic display, image splitter autostereoscopic display and dual-panel autostereoscopic display.
Contrast discrimination is an important type of information for establishing image quality metrics based on human vision. We used a dual-masking paradigm to study how contrast discrimination can be influenced by the presence of adjacent stimuli. In a dual masking paradigm, the observer's task is to detect a target superimposed on a pedestal in the presence of flankers. The flankers (1) reduce the target threshold at zero pedestal contrast; (2) reduce the size of pedestal facilitation at low pedestal contrasts; and (3) shift the TvC (Target threshold vs. pedestal contrast) function horizontally to the left on a log-log plot at high pedestal contrasts. The horizontal shift at high pedestal contrasts suggests that the flanker effect is a multiplicative factor that cannot be explained by previous models of contrast discrimination. We extended a divisive inhibition model of contrast discrimination by implementing the flanker effect as a multiplicative sensitivity modulation factor that account for the data well.
Spatial resolution is one of the main characteristics of electronic imaging devices such as the digital still-picture camera. It describes the capability of a device to resolve the spatial details of an image formed by the incoming optical information. The overall resolving capability is of great interest although there are various factors, contributed by camera components and signal processing algorithms, affecting the spatial resolution. The spatial frequency response (SFR), analogous to the MTF of an optical imaging system, is one of the four measurements for analysis of spatial resolution defined in ISO/FDIS 12233, and it provides a complete profile of the spatial response of digital still-picture cameras. In that document, a test chart is employed to estimate the spatial resolving capability. The calculations of SFR were conducted by using the slanted edge method in which a scene with a black-to- white or white-to-black edge tilted at a specified angle is captured. An algorithm is used to find the line spread function as well as the SFR. We will present a modified algorithm in which no prior information of the angle of the tilted black-to-white edge is needed. The tilted angle was estimated by assuming that a region around the center of the transition between black and white regions is linear. At a tilted angle of 8 degree the minimum estimation error is about 3%. The advantages of the modified slanted edge method are high accuracy, flexible use, and low cost.
In this paper, the techniques as well as the measurement results of the performance of commercial digital still- picture cameras are presented. The key parameters such as the camera Opto-Electronic Conversion Function (OECF), the noise based ISO speed, and the spatial frequency response (SFR) are reported. The camera OECF is defined as the relationship between the input luminance and the grayscale or digital output from the camera, which was measured by using a test chart with twelve squares of various luminances. The ISO speed was calculated from the exposure time, the effective f-number, and the luminance at different incremental signal-to-noise ratios. In general, the exposure time is not obtainable from a commercial digital camera unless a destructive measurement is undergoing. In this study, a device was setup to obtain the exposure time when the OECF test chart was recording. A modified slanted-edge method was employed to estimate the SFR by imaging a pattern with a black-to-white edge tilted at an arbitrary angle. There are seven digital still picture cameras as our test samples whose CCD sensor contains VGA-size and million pixels. The camera OECF of these cameras did not show significant difference under a large range of illumination. However, the ISO speed and the SFR were of great variation.
In this paper, we proposed an efficient algorithm to do image stabilization for digital camcorder. This approach is based on gray-coded bit-plane block matching to eliminate the unpleasing effect caused by involuntary movement of camera holders. To improve moving object detection and stabilization performance, a frame is divided into several blocks to do localized motion estimation. Based on our architecture, the temporal correlation is used at the motion unit to efficiently detect moving objects and panning conditions. To compensate for camera rotation, an energy minimization is also applied to calculate the coefficients of affine transform without many complicated computations. Having considered both programming flexibility and hardware efficiency, the motion decision and motion compensation units are coded in a microprocessor that interconnects with the stabilization hardware. The proposed stabilizer is now implemented on FPGA 10K 100.
The sharing and exchange of color images over the Internet pose very challenging problems to color science and technology. Emerging color standards will solve many of the problems we face today, but existing images of unknown origin and output devices of unknown calibration will continue to cause problems for many users. This paper presents a brief overview of available solutions to some of the problems and suggests some directions for future research. Although most existing solutions are quite [primitive and fragile, the rapid advance of computing technology promises to bring more sophisticated and intelligent image processing algorithms to common practical use. Image understanding, scene physics, visual calibration, and image perception are four areas of research that are beginning to make good progress toward a fully automatic quality optimization system for color imaging applications.
In this paper, we present a method to estimate the reflected and fluorescent spectral radiance factors of a fluorescent object based on spectrophotometric data without using a monochromator. We use truncated Fourier series to approximate both two spectral radiance factors. Then, based on the measured spectral obtained from a spectroradiometer, the coefficients of the truncated Fourier series are estimated using an weighted least square algorithm. The weighting function is defined as the sum of the CIE standard x, y, and z color matching functions. With the estimated reflected and fluorescent spectral radiance factors, we can predict the color appearance of a fluorescent object under other sources such that the color difference is minimized from viewpoint of human vision.
This paper presents the design and production of calibration targets for digital input color devices. By experimentally determined gamut of surface color, this study redesigns the aim values based on ISO/FDIS 12641 and to meet process specifications of Noritsu QSS23-HRCRT photographic printer with silver halide photography. The calibration target includes four components: a set of 144 color patches (3 levels in lightness and 4 levels in chroma at 12 different hue angles) within printing gamut, a neutral scale containing 22 steps based on visual perception, a set of C- M-Y-K-R-G-B dye scales showing characteristics of photographic materials, and a series of facial colors ranked by red. This research will describe the meaning of each element, the use of colorimetric mapping to CIELCH for each element, the conversion of these patch into a RGB-mode electronic image file, and how to control the processing of color photographic materials. And we proposed an approach of dynamic subgroup linear interpolation to achieve high process quality of manufacturing calibration targets and cost-down. Finally, statistic results revealed that 99% of the patches are within 10 delta Eab of the aim values specified in this study from long-term test and 99% of the patches in the manufacturing batch are within 5 delta Eab of the mean values from short-term test.
As the popularity of color peripheral devices grows, the problem of color inconsistency among color devices becomes more and more important. The ICC Profile, specified by the International Color consortium (ICC), is a reasonable solution to achieve color consistency. This paper describes an implementation of automatic scanner ICC Profile generator. The generator mainly consists of a reference- file-parser, a target-image-analyzer, and a profile- parameters-evaluator. The reference-file-parser retrieves the XYZ and Lab values of color patches from the ANSI CGATS.5 compliance file. The RGB values of color patches are automatically extracted from the scanned ANSI IT8.7/2 target by the target-image-analyzer. Finally, the parameters that are required for profile construction are generated using the profile-parameters-evaluator. Based on the test of 4 types of scanners, the average (Delta) E*ab of less than 3 is achieved.
By the progress of computer, computer peripherals such as color monitor and printer are often used to generate color image. However, cross media color reproduction by human perception is usually different. Basically, the influence factors are device calibration and characterization, viewing condition, device gamut and human psychology. In this thesis, a color reproduction system based on color appearance model and gamut mapping is proposed. It consists of four parts; device characterization, color management technique, color appearance model and gamut mapping.
In this paper, we suggest how to segment the face when there is the man under complex environment, extracts the features from segmented the image and proposes a effective recognition system using the discrete wavelet transform (DWT). This algorithm is proposed by following processes. First, two gray-level images is captured with 256 level of the size of 256 X 256 in constant illumination. We use a Gaussian filter to remove noise of input image and get a differential image between background and input image. Second, a mask is made from erosion and dilation process after binary of the differential image. Third, facial image is divided by projecting the mask into input image. Most characteristic information of human face is in eyebrow, eyes, nose and mouth. In the reason, the facial characteristics are detected after selecting local area including this area. Fourth, detecting the characteristic of segmented face image, edge is detected with Sobel operator. Then, eye area and the center of face are searched by using horizontal and vertical components of edge. Characteristic area consists of eyes, a nose, a mouth, eye brows and cheeks, is detected by searching the edge of the image. Finally, characteristic vectors are extracted from performing DWT of this characteristic area and are normalized it between +1 and -1. Normalized vectors is used with input vector of neural network. Simulation results show recognition rate of 100% about learned image and 92% about test image.
Four key issues in wavelet zero-tree based image coding are investigates and presented, there are (1) Fast wavelet transform that save 1/2 and 3/4 processing for 1D signal and 2D signals respectively. (2) The selection of the best wavelet filters that yields best performance (PSNR vs. Bit rate) for most common seen images. (3) Recommendation of number of wavelet scales (or frequencies) for image coding by experiments and analysis.
Iterated Transformation Theory-Based Coding suffers from very high computational complexity in encoding phase. This is due to its exhaustive search. In this paper, our proposed image coding algorithm preprocess an original image to subband segmentation image by wavelet transform before image coding to reduce encoding complexity. A similar block is searched by using the 24 block pattern codes which are coded by the edge information in the image block on the domain pool of the subband segmentation. As a result, numerical data shows that the encoding time of the proposed coding method can be reduced to 98.82% of that of Joaquin's method, while the loss in quality relative to the Jacquin's is about 0.28 dB in PSNR, which is visually negligible.
We are presently constructing `AROPET', a rotating PET scanner for imaging small animals. The design of the system has flexible geometry, using four detectors. Each detector is made of a position-sensitive PMTs (Hamamatsu R3941) coupled with 18 X 16 small individual BGO scintillator crystals of dimension 2.6 X 2.6 X 25 mm3. Animals can be imaged in two modes. One is similar to a gamma camera in which the detectors are stationary and a 2D planar projection imaging is obtained. This mode is used for initial characterization of the bio-distribution of tracers. In the other mode the detectors are rotated through 90 degree(s), and the diameter can be adjusted between 22 cm - 40 cm. This mode resembles a conventional 3D PET scan using a partial detector ring. Thirty-one tomographic images can be obtained after rebinning and reconstruction. The field of view is 51.3 mm (transaxial) by 45.6 mm (axial). The spatial resolution of the planar projection mode, and the results of the planar image of a phantom and the dynamical images of the bio-distribution of F18-FDG in a mouse are discussed.
The paper had studied the influence of compression ratio of foam on printing quality of ink cartridge. Fundamental properties of foam were first introduced as a review. Thus, some basic models for compression of foam could have been built up to deal with the physical effect of compression on the back pressure of ink cartridge. It might actually make difference on many aspects for an ink-jet printer. By means of experiments, several individual cases had been done which composed of different combinations for the compressed foams and color inks. It turned out that the printing quality might be much influenced in the cases. The results implied that the compression ratio of foam should be correctly chosen in order to yield best quality on the print. Finally, detailed analysis explained to make a suggestion for the choice. It's expected to be very helpful in the future design of ink cartridge with the foam.
The influence of back pressure of ink cartridge on regular operation of ink supply system was dealt with in the paper. In recent years, large ink supply system has developed at lot o increase the print usage of cartridge; however, the amount of back pressure in the cartridge may change very much when the cartridge would be equipped with ink supply system. It might further result in different printing behavior of cartridge in the printer. It's much of interest int he study. Therefore, experiments were done to determine what influence might occur in one specific test system. It was found that ink droplet, nozzle firing, and print quality were significantly influenced over the time of ink supply. The experimental result was helpful in future design of ink supply system.
To speed up the printing speed of an inkjet printer, the manufacturers normally focus on increasing the droplet frequency response. Hence, it has become a very important technique to measure the droplet frequency response of an inkjet printhead. A magneto-electric method is proposed to measure the droplet frequency response. The magneto-electric apparatus contains a metallic detecting plate and a magnetic ring with a gap of about 100 micrometers filled with a nonmagnetic insulating material. The magnetic ring itself is made of a high-permeability alloy consisting of about 78% nickel and 22% iron. When an ink drop jetted from a nozzle makes a contact with the metallic detecting plate, which is perpendicular to the nozzle plate of a printhead, a current is conducted through the detecting plate immediately, and detected as a portion of expected signal. The expected signal is then processed by a signal processing circuit for counting the number of jetted drops, and determining the maximum droplet frequency response of the inkjet printhead as a function of the driving frequency of an applied voltage across the printhead.
Amorphous, hydrogenated silicon carbide (a-SiC:H) deposited by plasma-enhanced-chemical-vapor deposition has been used as the most important film of passivation layers in a thermal ink-jet printhead. When the printhead was thermal- cycled from room temperature to about 400 degree(s)C, the a-SiC:H film is sustained by a variety of thermal and mechanical stresses that are detrimental to it's integrity. Thermal stress changes of a-SiC:H films were varied with different CH4/SiH4 gas ratios. Microstructure investigation was mainly achieved by FTIR technique. Less variation of the Si-H absorption bond causes less thermal stress change. Thin-film adhesion is an important problem in thermal ink- jet printhead between the Ta thin film and a-SiC:H films. A qualitative measure of film adhesion can be made with the scratch tester. The adhesive critical load and Ta coating failure modes on a-SiC:H were acquired to examine the film adhesion on these two investigated films. The adhesion depends on the nature of the interfacial region, which depends on the interactions between the depositing Ta thin film and the surface a-SiC:H films. An increased effective contact area in the interfacial region promotes a good adhesion.
The paper proposed a high resolution single-chip monolithic inkjet printhead by combination glowing of nozzle plate on the silicon substrate and anisotropic etching. Ink channels are defined by a sacrificed layer and etched through a mesh network by anisotropic etching. Silicon based channels are strong enough to against the attack of ink solution. Surface planarization is achieved by using PECVD deposited low stress dielectric film on top of the channels to seal mesh cavities. The heater elements are buried in a thin- sandwiched membrane and face down like a back shooter. Ink slot is formed by etching the substrate from the backside and channel is then connected with ink slot. Electrical forming nozzle plate will be done later. In this structure, no more another ink channel material is necessary since it is formed on Si substrate. By direct forming nozzle plate on chip, alignment for bonding nozzle plate and chip could be avoided to cost down.
At first, this paper introduces the structure of the design- image capturing system using the single-stripe pattern. At second, its operating principle is introduced, its mathematical model is established and the calibrating method for it is put forward. At last, its prototype is produced and calibrated.
As we approach the new millennium, the ongoing aim of human society is not only for promoting scientific technology but also creating new industries. To achieve this goal, each person in industry must recognize anew that the real meaning of science is to explore the absolute truth. It is also important that people recognize that there are unlimited matters which we humans do now yet know.