This paper describes a real-time image processing system for correction and enhancement of fluoroscopic (video X-ray) image data obtained from a large area, flat-panel, solid- state medical image sensor. The amorphous silicon sensor is 1536 X 1920 pixels, measuring 20 X 25 cm; for operation at 30 frames per second, the real pixel data rate is approximately 45 MB/sec.

In this paper we present the study and the implementation of an optimized chain of segmentation operators. We implemented this chain in real time, consisting of a Deriche contour detection, double threshold, closing of contours and finally region labeling, on a multi-FPGA architecture. This architecture has four processing FPGAs and four memory modules. Deriche operator, closing of contours and labeling occupy each one an FPGA. Double threshold and detection of the extremities filled partially the forth FPGA. The slowest component of the chain is Deriche operator which can go up to 11.4 Mhz, assuring the process of an image every 40 ms. Deriche operator tries to extract the contours by assuming that a contour is a step super positioned by a white gaussian noise. Our implementation consists of a smoothing part of four second order filters and a Sobel as a derivation part. The second order filters are causal and non-causal horizontal and vertical operators. The gradient image passes through a double threshold filter to select the real contours and the crests and the background pixels. Closing of contours eliminates the false crests and finally the labeling gives a unique label to each closed region. The latency of the chain is in the order of three images. This implementation shows the efficiency of the chain and also it demonstrates the capabilities of our architecture as a prototyping system.

Existing techniques have been limited in terms of throughput and classification capabilities. This paper presents an innovative approach of handling the fastest off line steel mill speeds while surface imaging and defect analysis are done using standard technologies. This involves the use of consumer electronics technology such as CCD-cameras and multimedia PCs. Advances in PC technology, such as MMX, have made it possible to do all image processing in software versus the use of dedicated hardware. Among the topics to be presented are the system architecture made up of arrays of inexpensive cameras and the corresponding processing units. Innovative illumination concepts guarantee outstanding image quality at any line speed. Intelligent pre-processing has been introduced to compensate various environmental effects. Sophisticated defect analysis provides highest quality data for the final classification step. As a result, a 3000 ft./min steel surface inspection system with resolutions of 0.04' is indeed a 100% software solution.

An acid-sensitive color-changing recording material composed of halogenated copolymer and acid indicators has been developed. To improve its image quality the imaging processes and related photochemical reactions and the image stability are studied and a new type of recording films with high resolution and sensitivity developed. It can be used for real-time image recording and displaying for large screen displaying, microfilm recording, and for source diagnostics of UV-rays, electron and laser beam and as radiation dosimeters etc.

A real time contrast enhancement system utilizing histogram- based algorithms has been developed to operate on standard composite video signals. This low-cost DSP based system is designed with fixed-point algorithms and an off-chip look up table (LUT) to reduce the cost considerably over other contemporary approaches. This paper describes several real- time contrast enhancing systems advanced at the Sarnoff Corporation for high-speed visible and infrared cameras. The fixed-point enhancer was derived from these high performance cameras. The enhancer digitizes analog video and spatially subsamples the stream to qualify the scene's luminance. Simultaneously, the video is streamed through a LUT that has been programmed with the previous calculation. Reducing division operations by subsampling reduces calculation- cycles and also allows the processor to be used with cameras of nominal resolutions. All values are written to the LUT during blanking so no frames are lost. The enhancer measures 13 cm X 6.4 cm X 3.2 cm, operates off 9 VAC and consumes 12 W. This processor is small and inexpensive enough to be mounted with field deployed security cameras and can be used for surveillance, video forensics and real- time medical imaging.

Many automated processing systems used in semiconductor manufacturing perform an alignment of one pattern to another, or find the relationship between two coordinate systems by using pairs of measurements of points in both systems. This operation is also performed by other applications in robotics and image processing, such as the hand-eye transform and the stereo model for 3D-point estimation. This paper discusses the alignment and coordinate transform processes and the least squares criteria used in finding the best rotation, translation, and scale change for matching two point sets.

In this paper, the principle of range-image based on image coding is introduced. According to the encoded images obtained by our laser scan-based image-encoder, the algorithms are presented to produce the threshold value, turn the encoded images into binary-state, calculate the scanning angle and find the range image. The corresponding utility programs are completed, and verified by our experiments.

2D cellular automata have recently been applied to detect geometrical shape regions. In this paper, 2D cellular automata rules are given that detect rectangular (square) and triangular regions in the plane. The cellular automata presented here are extremely fast with detection results after one or two applications of the rules.

The paper presents the novel principle on constructing a new class of highly parallel fast stable numerical algorithms of linear algebra. In accordance with suggested principle the parallel modifications of classic Gram-Shmidt algorithm and the conjugate directions' algorithm are developed. The modified algorithms have super-linear convergence rate: the sufficient number of iterations is proportional to effective rank of linear algebraic system. Also they have high numerical stability supported by easy controlled parallel cross-feedbacks. The new algorithms provide for high computing density in distributed processors and proposed for using in area of multidimensional data stream real-time processing.

Most image analysis/understanding applications require accurate computation of camera motion parameters. However, in multimedia applications, particularly in video parsing, the exact camera motion parameters such as the panning and/or zooming rates are not needed. The detection--i.e., a binary decision--of camera motion is all that is required to avoid declaring a false scene change. As camera motions can induce false scene changes for video parsing algorithms, we propose a fast algorithm to detect such camera motions: camera zoom and pan. As the algorithm is only expected produce a binary decision, without the exact panning/zooming rates, the proposed algorithm runs on a reduced data set, namely the projection data. The algorithm begins with a central portion of the image and computes the projection data (or the line integrals along the x- or y-axis) to turn the 2D image data into a 1D data. This projected 1D data is further processed via correlation processing to detect camera zoom and pan. Working with projection data saves processing time by an order of magnitude, since for instance, a 2D correlation takes N² multiplies per point, however a 1D correlation takes N multiplies per point. The efficacy of the proposed algorithm is tested for a number of image sequences and the algorithm is shown to be successful in detecting camera motions. The proposed algorithm is expected to be beneficial for video parsers working with Motion-JPEG data stream where motion vectors are not available.

Simulation of special effects such as: defocus effect, depth-of-field effect, raindrops or water film falling on the windshield, may be very useful in visual simulators and in all computer graphics applications that need realistic images of outdoor scenery. Those effects are especially important in rendering poor visibility conditions in flight and driving simulators, but can also be applied, for example, in composing computer graphics and video sequences- -i.e. in Augmented Reality systems. This paper proposes a new approach to the rendering of those optical effects by iterative adaptive filtering using spatial convolution. The advantage of this solution is that the adaptive convolution can be done in real-time by existing hardware. Optical effects mentioned above can be introduced into the image computed using conventional camera model by applying to the intensity of each pixel the convolution filter having an appropriate point spread function. The algorithms described in this paper can be easily implemented int the visualization pipeline--the final effect may be obtained by iterative filtering using a single hardware convolution filter or with the pipeline composed of identical 3 X 3 filters placed as the stages of this pipeline. Another advantage of the proposed solution is that the extension based on proposed algorithm can be added to the existing rendering systems as a final stage of the visualization pipeline.

Video transmission in a local area network consumes a great deal of network resources. Hybrid Analog/Digital Video Networks (ADViNet) can remedy this deficiency. ADViNet allows users to send real-time full-screen motion images through an analog communication medium without slowing down regular network activities. In this paper we are going to consider a multilayer control and management protocol that will allow users to supervise the analog medium through the digital one. Several groups of new protocol commands will be introduced. The first group covers the operations that directly change the ways of distributing video information through the analog medium. The second group includes network administration commands. Some of the ADViNet functions are intelligent: besides data transmission, they include data processing and/or decision-making elements.

The theory and method of computer simulations created formerly permits to investigate the interaction of high power laser radiation up to gigawatt on centimeter to the second power with the atoms as well as with small so with large fine structure interval in the external magnetic field. The modifications of the energies and populations of atomic levels due to such interaction lead to the new nonlinear effects. Among them besides the modifications in Raman effect and new effects of rotation of plane of polarization and circular dichroism investigated before there are such new effects as the modifications of the number, intensities, polarizations of the lines at the excitation spectra and the scattering spectra of resonant fluorescence. The real atom is multilevel system. Usually the theoretical investigations of nonlinear resonant magnetooptical effects are limited by two- or three-level approximations or by using the perturbation theory. This is justified for considering the effects in the weak magnetic fields but not sufficient for revealing and investigation the characteristics of the new nonlinear magnetooptical effects taking place in the intense radiation and magnetic fields. Such effects are related to modifications of electronic structure of atoms (level shifts and splittings, populations modifications). As the considered media is atomic gas one can reach for power density up to gigawatt on centimeter to the second power. As the considered concentrations are small it is possible to neglect the collisional relaxation. For real atom the considered system is multilevel so only numerical solution is appropriated. The simulations permit to find some new effects in the excitation spectra of Raman effect and resonant fluorescence.