We present a number of methods to detect and track multiple moving biological objects in image sequences acquired by different imaging techniques coupled to video microscopy. Movement and motility analysis is an important topic in biology and it is of major importance to be able to analyze the image sequences in order to get reliable and reproducible quantitative data such as number, position, movement phases and speed of the biological objects, as this information helps to characterize the biological assays. The detection is automatic and, in the case of phase contrast microscopy, is based upon the correlation of the image with a filter which varies adaptively to represent an object as it moves and deforms; in fluorescent imaging, the automatic detection is based on thresholding and mathematical morphology to determine and select the objects. The tracking is performed using a Kalman filter and a cost function which enable the position of the moving objects to be predicted, refined and updated. Once all moving objects have been assigned with unique spatio-temporal paths, trajectories are analyzed in terms of different parameters relevant for the motion analysis of biological objects.

Automatic object segmentation in highly noisy image sequences, composed by a translating object over a background having a different motion, is achieved through joint motion-texture analysis. Local motion and/or texture is characterized by the energy of the local spatio-temporal spectrum, as different textures undergoing different translational motions display distinctive features in their 3D (x,y,t) spectra. Measurements of local spectrum energy are obtained using a bank of directional 3rd order Gaussian derivative filters in a multiresolution pyramid in space- time (10 directions, 3 resolution levels). These 30 energy measurements form a feature vector describing texture-motion for every pixel in the sequence. To improve discrimination capability and reduce computational cost, we automatically select those 4 features (channels) that best discriminate object from background, under the assumptions that the object is smaller than the background and has a different velocity or texture. In this way we reject features irrelevant or dominated by noise, that could yield wrong segmentation results. This method has been successfully applied to sequences with extremely low visibility and for objects that are even invisible for the eye in absence of motion.

This paper describes a new fiducial detection method for use under varying lighting conditions without manual control of any parameters. We developed the algorithm especially for vision-based Augmented Reality (AR) systems. The major problem in AR is the registration between the virtual world and the real world. The user's pose in both worlds should be exactly the same. Vision-based AR is an attractive approach to the registration problem, however the fiducial detection methods used in many systems operate only under restricted lighting conditions. We developed a rule-based algorithm to segment regions of an image to detect known fiducials under varying lighting conditions. The algorithm is based on simple spatial and intensity relations among fiducials and their backgrounds. Rules and membership functions are defined from those relations. Rules are applied to find transition regions, and membership functions locate an edge position within a transition region. Edges are clustered to segment regions in an image. A vision-based AR system using our method operates under varying lighting conditions, including uneven lighting. This detection method extends the operating conditions of vision-based AR systems.

Image sequence quality enhancement is required in many different application areas, like restoration of old, damaged films, or displaying compressed sequences in higher spatial and/or temporal resolution using MC (motion- compensated) interpolation. Image sequence interpolation is the process of increasing the frame rate of a video signal by computing intermediate images between two or more known ones. It is used in a wide range of applications from low bitrate video conferencing till standard format conversion. The pixel values of the unknown frames have to be interpolated along the motion trajectories. First, correspondence must be established between the known images using a motion estimation algorithm, than this motion information is used to compute the interpolated image (or images) between the known ones. Interpolation of pixel values is done by a linear interpolation filter along the calculated motion trajectories. In our contribution experimental comparison of different motion models is given using artificial image sequences. The sequences have been generated by moving natural images along different trajectories. In the experiments presented in this paper the motion parameters are calculated from the known trajectories, and these parameters are passed to the interpolation algorithm. In other experiments, using real- life sequences, motion estimation should be used. It is in our case a multiresolution pel-recursive motion estimation algorithm.

This paper deals with the estimation of a dense displacement vector field between two successive images in a sequence using the markovian modelization. The optical flow is processed in two different hierarchical frameworks with the regularized constraint model we proposed. This model is derived from the potential function introduced by Geman and MacClure, in which we integrate the local motion amplitude obtained by Fourier analysis. It enables adaptive smoothness and then preserves motions discontinuities. First we apply a coarse-to-fine strategy in a standard multiresolution pyramid. We use the ICM algorithm only on the finest resolution scale of the pyramid, and the simulated annealing on the other scales. Secondly, we work with the multiscale scheme which allows only one resolution for the observations and a pyramidal structure for the primitives (the estimated optical flow). The results obtained on synthetic and real images sequences show that the estimation is efficiency increased. In our second contribution in this paper, we define a criterion for the determination of the regularization hyperparameter which controls the weight of the regularization term in the energy function. It is based on the entropy of the estimated motion vectors. An experimental study of this entropy allows to find the value of the hyperparameter which best fits a given sequence.

This paper proposes to use a multiresolutional spatio- temporal metric for the segmentation of an image sequence. In particular, scene-cut detection performance from an image sequence will be furnished. Wavelet decomposition is used for the multiresolutional analysis. The segmentation results obtained here can be used in the video browsing and indexing in multimedia applications.

The system presented here enhances documentation and data- secured, second-opinion facilities by integrating video into DICOM3.0. Digital stereoscopic video sequences (DSVS) are especially in demand for surgery (laparoscopy, microsurgery, surgical microscopy, second opinion, virtual reality). Therefore DSVS are also integrated into the DICOM video concept. We present an implementation for a medical video server extended by a DICOM interface. Security mechanisms conforming with DICOM are integrated to enable secure internet access. Digital (stereoscopic) video sequences relevant for surgery should be examined regarding the clip length necessary for diagnosis and documentation and the clip size manageable with today's hardware. Methods for DSVS compression are described, implemented, and tested. Image sources relevant for this paper include, among others, a stereoscopic laparoscope and a monoscopic endoscope. Additionally, an approach is presented to analyze the motion of the endoscopic camera for future automatic video- cutting.

Inherent in Kodak's solid-state high-speed video recording and playback systems is the ability to randomly access any given pixel within a 4D volume. Specifically, this refers to the ability to access a pixel of any given (x.y) coordinate of a given frame time t and of a given recording session s. A new set of techniques is presented for the rapid evaluation of entire video sequences which takes advantage of this wealth of video data. This technique is called `streak visualization'. As implied in the name, streak visualization combines the analytical compactness of a streak image with the versatility and interactivity implied by a visualization tool. Streak visualization generally involves the creation of appropriate 2D surfaces which slices the data volume in such a way as to reveal useful information. As will be demonstrated through a few illustrative examples, there are many possible applications of this technique.

An imaging camera that is used with x-ray radiography systems in high explosive experiments has been built and fielded. The camera uses a 40 mm diameter Micro-Channel Plate Intensifier (MCPI) for optical gain and photographic film for image recording. In the normal location of the x- ray film pack, a scintillating screen is placed instead. The camera system views the screen and records the image. The sensitivity of the MCPI to light makes the camera design sensitive to small details that a film pack does not need to consider. The x-ray image recording system was designed and built for situations where the film pack of the x-ray shadowgraph is not retrievable after the experiment. The system has been used in a number of experiments.

A repetitively pulsed broad band visible illumination system has been developed that is suitable for capturing images of high speed motion over sizable areas. At full pulse energy, a two lamp system can illuminate 60 square feet for movies at f/4 with 400 ASA color film and framing rates as high as 1700 fps. At reduced energy, for smaller area applications, the framing rate can be doubled. The short pulse length (4.5 microsecond(s) at full energy, 1.5 microsecond(s) at reduced energy) produces sharp images of high speed objects. This paper reports developments since the last presentation, including: (1) higher pulse repetition rates (a few kilohertz), (2) synchronization with high speed camera, (3) full scale burst of several thousand pulses, (4) characteristics of a compact demonstration system, and (5) demonstration of the ability of the short pulse to freeze motion.

This paper reports on the progress made in developing a practical illumination system to replace argon candles. The basic concept--inexpensive, disposable, modular, surface discharge lamps that provide a flexible illumination geometry and which can be sized to match the impedance of the pulse forming network (PFN)--was described in the previous paper. Since then, the pulse shape and energy output have been measured for the complete system for the three different 19 kJ PFNs, having characteristics times of 70 microsecond(s) , 200 microsecond(s) , and 1 ms. Exposure tests confirmed that the illumination level is sufficient for color photography at 2 Mfps with a Cordin 330A and for 10 ns exposures with an IMACON 468. As the illumination source for plate fragmentation tests using live explosives, the system created `high quality images' and proved to be a useful tool that was easily integrated into the test procedure. The laboratory demonstration system has since been converted into a production model that has been installed in an instrumentation van at Eglin AFB to provide illumination at various ranges.

The ISIS, In-situ Storage Image Sensor, may achieve the frame rate higher than 1,000,000 pps. Technical targets in development of the ISIS are listed up. A layout of the ISIS is presented, which covers the major targets, by employing slanted CCD storage and amplified CMOS readout. The layout has two different sets of orthogonal axis systems: one is mechanical and the other functional. Photodiodes, CCD registers and all the gates are designed parallel to the mechanical axis systems. The squares on which pixels are placed form the functional axis system. The axis systems are inclined to each other. To reproduce a moving image, at least fifty consecutive images are necessary for ten-second replay at 5 pps. The inclined design inlays the straight CCD storage registers for more than fifty images in the photo- receptive area of the sensor. The amplified CMOS readout circuits built in all the pixels eliminate line defects in reproduced images, which are inherent to CCD image sensors. FPN (Fixed Pattern Noise) introduced by the individual amplification is easily suppressed by digital post image processing, which is commonly employed in scientific and engineering applications. The yield rate is significantly improved by the elimination of the line defects.

An integrated, high-speed photographic system combining a high-repetition rate, pulsed ruby laser and a high-framing rate CCD camera has been demonstrated. Individually, the laser and camera have been discussed previously and each was developed under the Small Business Innovative Research sponsorship through the Air Force Research Lab. This paper presents for the first time digital images captured at 333 kHz using the two elements integrated as a high-speed photographic system.

A variety of range gated imaging experiments using high- speed gated/shuttered proximity focused microchannel plate image intensifiers (MCPII) are reported. Range gated imaging experiments were conducted in water for detection of submerged mines in controlled turbidity tank test and in sea water for the Naval Coastal Sea Command/U.S. Marine Corps. Field experiments have been conducted consisting of kilometer range imaging of resolution targets and military vehicles in atmosphere at Eglin Air Force Base for the U.S. Air Force, and similar imaging experiments, but in smoke environment, at Redstone Arsenal for the U.S. Army Aviation and Missile Command. Wavelength of the illumination laser was 532 nm with pulse widths ranging from 6 to 12 ns and comparable gate widths. These tests have shown depth resolution in the tens of centimeters range from time phasing reflected LADAR images with MCPII shutter opening.

Optical methods for testing of fluid flow are usually referred as non-intrusive or non-disturbing. Although modern optics is rapidly developing, new methods are rather sophisticated and cannot be easily adopted, which sometimes lead to enormous sum of money paid for technical equipment. There are various methods on how to study the interaction of light with fluid flow, one of which is the Schlieren method. Schlieren systems are based on the refraction of light rays when travel from one medium to another. The schlieren system used in this paper is called the Positive-Negative Grid schlieren system, a modification of R. Burton's schlieren system (1948). Simple relationships between density, temperature and light are investigated. Based on the relationships, a mathematical model for converting schliere data into temperature distribution can be constructed. This paper focuses on the thermal properties that could be extracted from schlieren images, using the model proposed. Problems with obtaining accurate temperature data from schlieren images are normally due to the noise in the images and heat radiation from fluid flow, which alters the schliere data slightly.

We present a wavelet based multiresolution extended Kalman filter (EKF) reconstruction approach to curved ray optical tomography. A state variable model describing the tomographic process is set up, and an EKF is applied to the wavelet transformed model to estimate the refractive index distribution of an optically transparent refracting object from noisy optical path-length difference (OPD) data. Preliminary results of reconstructions of a synthetic time- invariant refractive index distribution from OPD data sets of various noise levels are comparable with those obtained from a typically used deterministic approach, the average correction per projection method.