There has been quite a bit of discussion lately regarding the worth of research universities and their value to education and society in general. I would like to enter the fray with some down-home logic...

Real-time image processing is any image processing where timeliness is as critical as accuracy. Included in this domain are real-time image compression, target acquisition and tracking, remote control and sensing, image enhancement and filtering, networking for realtime imaging, advanced computer architectures, computer vision, optical measurement and inspection, and simulation...

A real-time application of computer vision concerning tracking and inspection of a submarine pipeline is described. The objective is to develop automatic procedures for supporting human operators in the real-time analysis of images acquired by means of cameras mounted on underwater remotely operated vehicles (ROV). Implementation of such procedures gives rise to a human-machine system for underwater pipeline inspection that can automatically detect and signal the presence of the pipe, of its structural or accessory elements, and of dangerous or alien objects in its neighborhood. The possibility of modifying the image acquisition rate in the simulations performed on video-recorded images is used to prove that the system performs all necessary processing with an acceptable robustness working in real-time up to a speed of about 2.5 kn, widely greater than that the actual ROVs and the security features allow.

Document image processing has become an increasingly important technology in the automation of office documentation tasks. Automatic document scanners such as text readers and optical character recognition (OCR) systems are an essential component of systems capable of those tasks. One of the problems in this field is that the document to be read is not always placed correctly on a flat-bed scanner. This means that the document may be skewed on the scanner bed, resulting in a skewed image. This skew has a detrimental effect on document analysis, document understanding, and character segmentation and recognition. Consequently, detecting the skew of a document image and correcting it are important issues in realizing a practical document reader. We describe a new algorithm for skew detection. We then compare the performance and results of this skew detection algorithm to other published methods from O’Gorman, Hinds, Le, Baird, Postl, and Akiyama. Finally, we discuss the theory of skew detection and the different approaches taken to solve the problem of skew in documents. The skew correction algorithm we propose has been shown to be extremely fast, with run times averaging under 0.25 CPU seconds to calculate the angle on a DEC 5000/20 workstation.

Real-time counting of pedestrians traveling through a transport system are increasingly required for traffic control and management by the companies operating such systems. One of the most widely used systems for counting passengers consists of a mechanical gate equipped with a counter. Such simple systems, however, are not able to count passengers jumping above the gates. Moreover, passengers carrying large luggage or bags may meet some difficulties when going through such gates. The ideal solution is a contact-free counting system that would bring more comfort of use for the passengers. For these reasons, we propose to use a video processing system instead of these mechanical gates. The optical sensors discussed offer several advantages, including well-defined detection areas, fast response time, and reliable counting capability. A new technology is developed and tested, based on linear cameras. For the algorithms, thanks to the principle of our system, no assumption is made about the scene being analyzed and the nature of pedestrian movements to enable the system to run in real time. We also consider the problems presented by crowded scenes, when a high incidence of pedestrians occlusions occurs. Preliminary results have shown that this system is very efficient when the passengers crossing the optical gate are well separated. In other cases, such as in compact crowd conditions, good accuracy in terms of counting in real time is demonstrated. These results are illustrated by means of a number of sequences shot in field conditions.

An adaptive scaled mean square error (SMSE) filter using a Hopfield neural-network-based algorithm is presented. We show the development of the original SMSE filter from the minimum mean square error (MMSE) filter and the parametric mean square error (PMSE) filter, both of which suffer from the oversmooth phenomena. The SMSE filter is more efficient than the PMSE filter in terms of noise removal as it does not take into account all the correlation factors used for image enhancement. To further improve the performance of the SMSE filter, an adaptive approach is introduced. The adaptive SMSE filter uses a mask operation technique. A user-defined mask is moved across the image and the filtering parameters are computed based on the local image statistics of the region below the mask. The original and the adaptive SMSE filters are implemented using a Hopfield neural-network-based algorithm. A number of experiments were performed to test the filter characteristics.

Real-time imaging has application in areas such as multimedia, virtual reality, medical imaging, and remote sensing and control. Recently, the imaging community has witnessed a tremendous growth in research and new ideas in these areas. To lend structure to this growth, we outline a classification scheme and provide an overview of current research in real-time imaging. For convenience, we have categorized references by research area and application.

A set theoretic approach for the spectral characterization of a color scanner is described. These devices usually employ three channels to obtain a device dependent RGB (red, green, blue) image. To display/print an image, the device dependent RGB values must be correctly transformed to the color space of the target device. To determine accurate and efficient transformations for a number of devices, knowledge of the spectral sensitivity of the scanneris essential. Direct measurement of the sensitivity requires a set of expensive narrow band reflectances and is often infeasible. Methods that estimate the spectral sensitivity based on measurements with typical reflectance samples are therefore of interest. Due to the low dimensionality of the space of object reflectance spectra, this is a highly ill-conditioned problem. As a result, conventional estimation techniques that fail to take a priori knowledge into account perform rather poorly on this problem. A set theoretic approach that incorporates available a priori knowledge into the estimation framework yields better results. Results are presented for a simulated scanner characterization problem and for an actual characterization to demonstrate the increased accuracy compared with conventional methods.

Temporal noise-reduction filtering of image sequences is commonly applied in medical imaging and other applications, and a common assessment technique is to measure the reduction in display noise variance. Theoretically and experimentally, we demonstrate that this is inadequate because it does not account for the interaction with the human observer. Using a new forced-choice method, we compare detectability of low-contrast objects and find a noise level for an unfiltered sequence that gives the same detectability as the filtered sequence. We report the equivalent detectability noise variance ratio, or EDVR. For a digital low-pass filter that reduces the bandwidth by 1/2, display noise reduction predicts an EDVR of 0.5. The measured value averaged over three subjects, 0.93±0.19, compares favorably with the 0.85 predicted from a theoretical human observer model, and both are very close to the value of 1.0 expected for no filtering. Hence, the effective, perceived noise is relatively unchanged by temporal low-pass filtering. The computational observer model successfully evaluates a simple low-pass temporal filter, and we anticipate that it can be used to predict the observer response to other image enhancement filters.

The conventional method for sending halftone images via facsimile machines is inefficient. The previously proposed ToneFac algorithm improves the transmission of halftone images. ToneFac represents a halftone image by mean gray values of the disjoint blocks and an error image, which records the difference between the desired halftone and the halftone generated using the mean gray values. To improve on ToneFac, we propose additional processing techniques: searching for the error-minimizing gray value for each block; quantization and coding of block values; bit switching, which transforms the error image into a more compressible image; optimal block sizing; and spurious dot filtering, which removes perceptually insignificant dots. The new algorithm is compared to other methods, including adaptive arithmetic coding, and is shown to provide improvement in bit rate. A theoretical consideration of the compression ratio from the ToneFac algorithm is also given.

Reviewed by A. Tanju Erdem Bilkent University Electrical and Electronics Engineering Department TR-06533 Bilkent, Ankara, Turkey