Military aircraft fly below 100 ft. above ground level in support of their missions. These aircraft include fixed and rotary wing and may be manned or unmanned. Flying at these low altitudes presents a safety hazard to the aircrew and aircraft, due to the occurrences of obstacles within the aircraft's flight path. The pilot must rely on eyesight and in some cases, infrared sensors to see obstacles. Many conditions can exacerbate visibility creating a situation in which obstacles are essentially invisible, creating a safety hazard, even to an alerted aircrew. Numerous catastrophic accidents have occurred in which aircraft have collided with undetected obstacles. Accidents of this type continue to be a problem for low flying military and commercial aircraft. Unmanned Aerial Vehicles (UAVs) have the same problem, whether operating autonomously or under control of a ground operator. Boeing-SVS has designed a passive, small, low- cost (under $100k) gimbaled, infrared imaging based system with advanced obstacle detection algorithms. Obstacles are detected in the infrared band, and linear features are analyzed by innovative cellular automata based software. These algorithms perform detection and location of sub-pixel linear features. The detection of the obstacles is performed on a frame by frame basis, in real time. Processed images are presented to the aircrew on their display as color enhanced features. The system has been designed such that the detected obstacles are displayed to the aircrew in sufficient time to react and maneuver the aircraft to safety. A patent for this system is on file with the US patent office, and all material herein should be treated accordingly.

Dermatologist as well as the cosmetical industry are interested in evaluating the mechanical properties of human skin. Many devices have been developed to measure skin's response to mechanical stress. In the presented paper a new approach to quantify the viscoelastic behavior of human skin on mechanical stress is proposed. Image processing techniques are used to detect the two-dimensional deformation of the skin in uniaxial tensile tests. The apparatus consists of a computer-controlled stepper motor drive mechanism to extend the skin, a load cell to measure displacement vector fields are calculated by a method based on local template matching and interpolation algorithms. From the displacement vector fields a strain tensor and the principal strain directions are evaluated. A model built up of springs and dashpots, is used to characterize the stress-strain-time relationships of skin and to obtain a set of parameters, which represent the instantaneous elasticity, the delayed elasticity and the viscosity of skin on loading. The results show the accuracy of the model. The method seems to be useful to investigate the influences of age, test area, cosmetics, etc. on the mechanical properties of human skin in vivo.

The aim of this application is to detect changes in an aerial scene by comparing stereo pairs taken at intervals of several years in order to update a database. The result is a set of image locations that have a high likelihood to contain changes. Each location will be submitted to a human operator who will either validate the given change and update the database or reject it. We are mainly interested in changes occurring for a specific class of objects : buildings. To isolate new construction, we provide an algorithm that works in two steps. First, during a focusing phase, we aim to eliminate a large part of the scene without losing any actual changes. This is achieved with a Digital Elevation Model (DEM) comparison between the two different dates. Then, in the second phase, we classify regions of interest (ROI). Each ROI is described by four images: a stereo pair of the focusing area at the first date and a stereo pair of the focusing area at the second date. To decide whether or not the ROI contains a change, we classify each of the four images as building or non-building. The building vs non-building classifier is a combination of several decision trees induced by learning. Each node of a decision tree is identified with a graph of features which is more likely to describe buildings than background. Finally, the classification results at the two different dates are compared.

Imagery is collected much faster and in significantly greater quantities today compared to a few years ago. Accurate registration of this imagery is vital for comparing the similarities and differences between multiple images. Image registration is a significant component in computer vision and other pattern recognition problems, medical applications such as Medical Resonance Images (MRI) and Positron Emission Tomography (PET), remotely sensed data for target location and identification, and super-resolution algorithms. Since human analysis is tedious and error prone for large data sets, we require an automatic, efficient, robust, and accurate method to register images. Wavelet transforms have proven useful for a variety of signal and image processing tasks. In our research, we present a fundamentally new wavelet-based registration algorithm utilizing redundant transforms and a masking process to suppress the adverse effects of noise and improve processing efficiency. The shift-invariant wavelet transform is applied in translation estimation and a new rotation-invariant polar wavelet transform is effectively utilized in rotation estimation. We demonstrate the robustness of these redundant wavelet transforms for the registration of two images (i.e., translating or rotating an input image to a reference image), but extensions to larger data sets are feasible. We compare the registration accuracy of our redundant wavelet transforms to the critically sampled discrete wavelet transform using the Daubechies wavelet to illustrate the power of our algorithm in the presence of significant additive white Gaussian noise and strongly translated or rotated images.

The human visual system (HVS) is highly non-uniform in sampling, coding, processing and understanding. The spatial resolution of the HVS is highest around the point of fixation (foveation point) and decreases rapidly with increasing eccentricity. Currently, most image quality measurement methods are designed for uniform resolution images. These methods do not correlate well with the perceived foveated image quality. Wavelet analysis delivers a convenient way to simultaneously examine localized spatial as well as frequency information. We developed a new image quality metric called foveated wavelet image quality index (FWQI) in the wavelet transform domain. FWQI considers multiple factors of the HVS, including the spatial variance of the contrast sensitivity function, the spatial variance of the local visual cut-off frequency, the variance of human visual sensitivity in different wavelet subbands, and the influence of the viewing distance on the display resolution and the HVS features. FWQI can be employed for foveated region of interest (ROI) image coding and quality enhancement. We show its effectiveness by using it as a guide for optimal bit assignment of an embedded foveated image coding system. The coding system demonstrates very good coding performance and scalability in terms of foveated objective as well as subjective quality measurement.

Fast algorithms for a wide class of non-separable n-dimensional (nD) discrete unitary K-transforms (DKT) are introduced. They need less 1D DKTs than in the case of the classical radix-2 FFT-type approach. The method utilizes a decomposition of the nD K-transform into the product of a new nD discrete Radon transform and of a set of parallel/independ 1D K-transforms. If the nD K-transform has a separable kernel (e.g., the case of the discrete Fourier transform) our approach leads to decrease of multiplicative complexity by the factor of n comparing to the classical row/column separable approach. It is well known that an n-th order Volterra filter of one dimensional signal can be evaluated by an appropriate nD linear convolution. This work describes new superfast algorithm for Volterra filtering. New approach is based on the superfast discrete Radon and Nussbaumer polynomial transforms.

Image sequences recorded from moving platforms are often distorted by undesired motion. The image motion within a video sequence captured from a moving platform is compounded in general from a smooth component such as camera panning and a high frequency component such as mechanical vibration or shocks. The undesired high frequency component of the motion may affect severely the perception of the image sequence content. This motion component causes two types of distortion. The first is the unexpected scene change caused by motion and vibration of the line-of-sight from frame to frame. The second effect is blur of each frame of the sequence due to frame motion during its exposure. In this work we propose an algorithm that eliminates the undesired component of the motion and increases the spatial resolution of each frame in the image sequence. The resolution is increased by deblurring the images and by utilizing information within adjacent frames with superresoluion restoration techniques.

In living bodies, the correct perceptual representation of size constancy requires that an object's size appear the same when it changes its location with respect to the observer. At the same time, it is necessary that objects at different locations appear to be the same size if they are. In order to do that, the perceptual system must recover from the stimuli impinging on the individual, from the light falling on the retina, a representation of the relative sizes of objects in the environment. Moreover, at the same time, image perception is related to another type of phenomena. It corresponds to the well known perceptual illusions. To analyze this facts, we propose a system based on a particular arrays of receptive points composed by optical fibers and dummy fibers. The structure is based on the first layers of the mammalians primary visual cortex. At that part of the brain, the neurons located at certain columns, respond to particular directions. This orientation changes in a systematic way as one moves across the cortical surface. In our case, the signals from the above-mentioned array are analyzed and information concerning orientation and size of a particular line is obtained. With this system, the Muelle-Lyer illusion has been studied and some rules to interpret why equal length objects give rise to different interpretations are presented.

This paper presents two different watermarking schemes devoted to protect video objects. The first presented scheme performs embedding and detection in the uncompressed domain. It has been developed to enable signature detection after object manipulations such as rotations, translations and VOL modifications. To achieve these requirements, the first scheme exploits the shape of the object using CPA analysis: a random sequence is transformed to fit the scale and the orientation of the object. The detection of the mark is performed applying an inverse transform and calculating a correlation between the random sequence and the transformed object. The second scheme is based on compressed-domain processing of video objects. Two different signals are embedded, one for synchronization recovery and another for copyright protection. Both signals are embedded and detected in the compressed domain. During detection, first synchronization recovery is performed and then the copyright watermark is extracted.

Transmission and exchange of digital images with friends and customers is become a very simple task thanks to the development of the communication networks and the tools built around them. Unfortunately, such operations become delicate whenever image security is required, typically for commercial applications or protection of proprietary data. Solutions associated to data encryption already exist but are usually complex and do not take into account the specificities of images, generally under a compressed form. The Joint Photographic Experts group has recently created a new still image coding standard called JPEG 2000. It presents an efficient compression scheme together with support of functionalities required by today and tomorrow applications (progressive decoding, region of interest...). By considering the JPEG 2000 algorithm, we are presenting tools for image authentication or access control (on image resolutions and qualities). These techniques can be applied on JPEG 2000 codestreams or directly integrated into the coding/decoding operations, and are mainly based on modification and insertion of information in the bit stream. The resulting codestreams remain compliant with the standard. Moreover, they allow image side information retrieval and/or prevent use by unauthorized parties.

There is a wide consensus among the feature film production studios that the Internet era brings a new paradigm for film distribution to cinemas worldwide. The benefits of digital cinema to both producers and cinemas are numerous: significantly lower distribution and maintenance costs, immediate access to film libraries, higher presentation quality, and strong potential for developing new business models. Despite these advantages, the studios are still reluctant to jump into the digital age. The main showstopper for digital cinema is the danger of widespread piracy. Piracy already costs Hollywood an estimated two billion dollars annually and digital cinema without proper copyright enforcement could increase this number. In this paper, we present a copyright management system that aims at providing the set of necessary security tools: standard cryptographic primitives and copyright protection mechanisms that enable a reliable and secure feature film delivery system.

This paper present the concept of robust video hashing as a tool for video identification. We present considerations and a technique for (i) extracting essential perceptual features from a moving image sequences and (ii) for identifying any sufficiently long unknown video segment by efficiently matching the hash value of the short segment with a large database of pre-computed hash values.

In this paper we introduce a new content-fragile watermarking concept for multimedia data authentication, especially for a/v data. While previous data authentication watermarking schemes address single media stream only, we discuss the requirements of multimedia protection techniques. Furthermore we introduce our new approach called 3D thumbnail cube. The main idea is based on a 3D hologram over continuing video and audio frames. Beside the data authentication, we face the owner authentication problem as second requirement for manipulation recognition. The watermark for manipulation recognition has to be created by the owner itself. The goal of owner authentication is to ensure that an entity is the one it claims to be. Therefore we introduce a key server and a biometric hash approach. We discuss several strategies and introduce a biometric based framework for owner authentication. With our presented data and owner authentication solutions we can realize an advanced security level.

With the advent of digital technologies, many new market opportunities have emerged for content owners, content distributors, and consumer electronics/information technology industries. An essential requirement for developing a thriving marketplace is the protection of copyrighted content in digital form. There are four major stages in the delivery of content to the consumer: (1) capturing on digital media, (2) packaging, (3) distribution to home networks, and (4) transfer to the final audio/visual device within the home network. Entertainment content is of particular importance as it will be in high demand for many years to come. If an end-to-end security cannot be provided in a digital market, there would be no incentive for content creation. Lack of new supplies would result in detrimental effects for all the industries involved in the delivery chain. In this paper, we present the primary means of securing the entertainment content from creation to consumption in an attempt to understand the overall complexity of the problem.

Watermarking can be modeled as a transmission through a steganographic channel. Most of the channels studied up to now in the literature were additive or substitutive channels where the noise modifies the data value. But other modifications may occur, as geometric distortions: images may be cropped, scaled, non-linearly distorted etc. We model such transformations by a geometric channel, where the noise modifies the geometry of the image rather than the data values. We study the jitter channel which copies and deletes randomly some lines of the image. We use a Markov chain to model the channel, and propose an algorithm to enhance the detection of a watermark embedded by a spread-spectrum technique subject to a jitter attack.

Although it is obvious that watermarking has become of great interest in protecting audio, videos, and still pictures, few work has been done considering 3D meshes. We propose a new method for watermarking 3D triangle meshes. This method embeds the watermark as triangles deformations. The list of watermarked triangles is obtained through a similar way to the one used in the TSPS (Triangle Strip Peeling Sequence) method. Unlike TSPS, our method is automatic and more secure. We also show that it is reversible.

This paper describes the functional model of a combined conditional access and fingerprinting copyright (-or projectionright) protection system in a digital cinema framework. In the cinema industry, a large part of early movie piracy comes from copies made in the theater itself with a camera. The evolution towards digital cinema broadcast enables watermark based fingerprinting protection systems. Besides an appropriate fingerprinting technology, a number of well defined security/cryptographic tools are integrated in order to guaranty the integrity of the whole system. The requirements are two-fold: On one side, we must ensure that the media content is only accessible at exhibition time (under specific authorization obtained after an ad-hoc film rental agreement) and contains the related exhibition fingerprint. At the other end, we must prove our ability to retrieve the fingerprint information from an illegal copy of the media.

Digital data representation provides an efficient and fast way to access to information and to exchange it. In many situations though ownership or copyright protection mechanisms are desired. For still images and video, one possible way to achieve this is through watermarking. Watermarking consists of an imperceptible information embedded within a given media. Parallel Processing Watermarking Embedding Schemes have demonstrated to be efficient from a computational and memory usage point of view for very large images. These schemes consist in dividing the image into tiles and watermarking each independently. The processing allows the use of a parallel computation scheme. The watermarking method used in the scope of this work is a parallel variant of an approach known as self-referenced Spread Spectrum signature pattern. Since the watermarking scheme has been modified through tiling, the extra references due to signature replication can be used in the retrieval. This work describes the above mentioned approach to watermark images and provides an analysis of its performance.

The JPEG 2000 image compression system offers significant opportunity to improve imaging over the Internet. The JPEG 2000 standard is ideally suited to the client/server architecture of the web. With only one compressed version stored, a server can transmit an image with the resolution, quality, size, and region custom specified by an individual client. It can also serve an interactive zoom and pan client application. All of these can be achieved without server side decoding while using only minimal server computation, storage, and bandwidth. This paper discusses some of the system issues involved in Internet imaging with JPEG 2000. The choices of the client, passing of control information, and the methods a server could use to serve the client requests are presented. These issues include use of JPEG 2000 encoding and the decoding options in the standard. Also, covered are some proposed techniques that are outside the existing standards.

JPEG2000 Part I provides a host of compression options and data ordering choices which enable powerful applications and create tremendous flexibility in the handling of still images. Part I, however, is restricted to handle multiple component images (with the exception of three-component images) a single component at a time. In general, Part I allows no exploitation of inter-component correlation that may exist. Part II introduces a robust multiple component transform capability which is applied prior to the Part I spatial wavelet decomposition and compression. This paper describes some of the multiple component transform capabilities in JPEG2000 Part II, including prediction, traditional decorrelation, wavelet transformations, and reversible integer transformations.

Synthetic Aperture Radar (SAR) imagery has traditionally posed a challenge to image compression algorithms operating at low to moderate bit-rates (0.25 to 1.0 bits per pixel), because SAR texture is typically reconstructed as smooth fields. This smooth reconstruction is visually objectionable and conceals information from interpreters, who are accustomed to analyzing textures and using texture to define the context of reflecting point targets and clusters. JPEG 2000 is emerging as a new international standard for wavelet-based image compression, and it too tends to reconstruct SAR texture as smooth fields when operating at low or moderate bit-rates. This characteristic of the new standard motivates an attempt to carry texture synthesis techniques proven on other compression algorithms over to JPEG 2000. This present effort demonstrates the value of root-mean-square (RMS) reconstruction, a technique previously proven on a proprietary codec, for improving the visually perceived quality of JPEG 2000 compressed SAR images. RMS reconstruction is found to be extremely useful for JPEG 2000, both for improving the quality of compressed SAR images and also for improving the visual appearance of compressed electro-optical (EO) imagery.

To overcome many drawbacks in the current JPEG standard for still image compression, a new standard, JPEG2000, is under development by the International Standard Organization. Embedded bit plane coding is the heart of the JPEG2000 encoder. This encoder is more complex and has significantly higher computational requirements compared to the entropy encoding in current JPEG standard. Because of the inherent bit-wise processing of the entropy encoder in JPEG2000, memory traffic is a substantial component in software implementation. However, in hardware implementation, the lookup tables can be mapped to logic gates and memory accesses for the state bit computation can be reduced significantly by careful design. In this paper, we present an efficient VLSI architecture for embedded bit-plane coding in JPEG2000 that reduces the number of memory accesses. To better understand the interaction of this architecture with the rest of the coder, we also present a system level architecture for efficient implementation of JPEG2000 in hardware.

With the increasing importance of heterogeneous networks and time-varying communication channels, such as packet-switched networks and wireless communications, fine scalability has become a highly desirable feature in both image and video coders. A single highly scalable bitstream can provide precise rate control for constant bitrate (CBR) traffic and accurate quality control for variable bitrate (VBR) traffic. In this paper, we propose two methods that provide constant quality video under buffer constraints. These methods can be used with all scalable coders. Experimental results using the Motion JPEG2000 coder demonstrate substantial benefits.

Region Of Interest (ROI) coding is one of the innovative functionalities supported by JPEG 2000, the new ISO/ITU-T still image coding standard developed by the Joint Photographic Experts Group (JPEG). It enables a non-uniform distribution of the image quality between a selected region (the ROI) and the rest of the image (background). This feature is obtained by scaling background coefficients in the wavelet domain. According to the ROI Maxshift method defined in JPEG 2000 part 1 (baseline algorithm), the background bit-planes are down-shifted below all ROI coefficients. In addition, the ROI can have any shape as the latter does not need to be transmitted to a JPEG 2000 decoder (no codestream overhead due to the coding of the shape). On the contrary, this method requires decoding of all ROI coefficients before accessing bit-planes of the background. Furthermore, it uses large shifting values that significantly increase the number of total bit-planes to encode. In JPEG 2000 part 2 (extensions), a generic (Scaling based) ROI coding has been included. This method supports any scaling values. In particular, it allows a rough control on both ROI and background qualities distributions in the codestream, but implies the derivation of a ROI bit-mask at the decoding side. This paper starts by providing some hints on how to choose an optimal Maxshift scaling value, as well as how to pad the ROI extra bits appearing during the shift operation. Then, it proposes an encoding algorithm that combines advantages of both ROI methods. This algorithm can be used by applications where visually lossless ROI*s is acceptable and is based on an extension of the Maxshift method to low scaling values. The generated codestreams remain compliant with the Maxshift decoding algorithm described in JPEG 2000 part 1, and can be consequently handled by any JPEG 2000 decoder.

JPEG 2000 Part 2 (extensions) contains a number of technologies that are of potential interest in remote sensing applications. These include arbitrary wavelet transforms, techniques to limit boundary artifacts in tiles, multiple component transforms, and trellis-coded quantization (TCQ). We are investigating the addition of these features to the low-memory (scan-based) implementation of JPEG 2000 Part 1. A scan-based implementation of TCQ has been realized and tested, with a very small performance loss as compared with the full image (frame-based) version. A proposed amendment to JPEG 2000 Part 2 will effect the syntax changes required to make scan-based TCQ compatible with the standard.

It is well known that tile boundary artefacts occur in lossy wavelet-based image coding. The base model of the JPEG2000 standard (ie JPEG2000 Part I) suffers from these artefacts, being a wavelet-based coding system. This paper analyses the tile boundary problems of JPEG2000 Part I and presents a novel method for reducing these tile boundary artefacts. This method has recently been adopted as part of the JPEG 2000 Verification Model 9.0 and as an addition to Part II of the JPEG2000 standard.

We give a brief tour of the emerging ITU-T H.26L videocoding standard. Like its predecessors, H.26L is being designed partly for videoconferencing applications; however, many other applications are considered to be within the scope of the design effort and tests indicate that H.26L is fully suitable for a very broad range of applications. Starting from a clean slate, H.26L is a powerful new design that is forward-looking in its performance and scope of applications. We compare H.26L to the highest capabilities of the most complex and newest versions of prior standards such as H.263 and MPEG-4, and provide benchmarks for its performance. Our analysis indicates that the draft H.26L standard offers compelling advantages over all existing video coding standards. It has the potential to redraw the landscape of consumer and enterprise video applications.

We consider the problem of designing variable length codes for large symbol sets. Highly structured codes such as the Golomb codes are often used to reduce the encoding and decoding complexity. This structure can limit the coding efficiency. We use the Golomb codes to construct low complexity codes with improved coding efficiency. The construction can be used to provide codes for adaptive coding.

With the recent development of the use of digital media for cytogenetic imaging applications, efficient compression techniques are highly desirable to accommodate the rapid growth of image data. This paper introduces a lossy to lossless coding technique for compression of multiplex fluorescence in situ hybridization (M-FISH) images, based on 3-D set partitioning in hierarchical trees (3-D SPIHT). Using a lifting-based integer wavelet decomposition, the 3-D SPIHT achieves both embedded coding and substantial improvement in lossless compression over the Lempel-Ziv (WinZip) coding which is the current method for archiving M-FISH images. The lossy compression performance of the 3-D SPIHT is also significantly better than that of the 2-D based JPEG-2000.

Chromosome analysis is an important procedure in clinical and cancer cytogenetics. Efficient image compression techniques are highly desired to accommodate the rapid growth in the use of digital media for archiving, storage and communication of chromosome spread images. In this paper, we propose a new method based on an important characteristic of these images: the regions of interest to cytogeneticists for evaluation and diagnosis are well determined and segmented. Such information is utilized to advantage in our compression algorithm, which combines lossless coding of chromosome regions with lossy-to-lossless coding of the remaining image parts. This is accomplished by first performing a differential operation on chromosome regions for decorrelation, followed by critically sampled integer wavelet transforms on these regions and the remaining image parts. A modified set partitioning in hierarchical trees (SPIHT) algorithm is then used to generate separate embedded bitstreams that allow continuous lossy-to-lossless compression of both chromosome regions and the rest of the image (although lossless coding of the former is commonly used in practice). Experiments on sample chromosome spread images indicate that the proposed approach significantly outperforms several reference compression schemes and the techniques currently employed in commercial systems.

A joint classification-compression scheme that provides the user with added capability to prioritize features of interest in the compression process is proposed. The dual compression system includes a primary unit for conventional coding of multispectral image set followed by an auxiliary unit to code the resulting error induced on pixel vectors that represent features of interest. This technique effectively allows features of interest in the scene to be coded at a relatively higher precision level than the nonessential features. Prioritized features are selected from a thematic map or directly specified by their unique spectral signatures. Using the specified spectral signatures of the prioritized features as endmembers, a modified linear spectral unmixing procedure is applied to the original data as well as the decoded data. The resulting two sets of concentration maps, which represent prioritized features before and after compression, are compared and the differences between them are coded via an auxiliary compression unit and transmitted to the receiver along with conventionally coded image set. At the receiver, the recovered differences are blended back into the decoded data for an enhanced restoration of the prioritized features. The utility of this approach is that it works with any multispectral compression scheme. This method has been applied to test imagery from various platforms including NOAA's AVHRR (1.1 km GSD), and LANDSAT 5 TM (30 m GSD), LANDSAT 5 MSS (79 m GSD).

This paper describes a new approach for the motion estimation problem of digital video based upon the lowest requirement for residual encoding. Most Encoders today use Mean Absolute Difference (MAD) or Mean Square Error (MSE) as the test for the best match. These tests do not give any meaningful measure of how much data may be needed to encode the residuals. Our technique is based on using a new criterion to determine the best match block. The new criteria is based on the minimum number of coefficients to represent the residuals. Experimental results show that, using the new proposed technique we achieve a higher compression ratio for the same quality.

The enhancement layer in many scalable coding algorithms is composed of residual coding information. There is another type of information that can be transmitted instead of (or in addition to) residual coding. Since the encoder has access to the original sequence, it can utilize adaptive format conversion (AFC) to generate the enhancement layer and transmit the different format conversion methods as enhancement data. This paper investigates the use of adaptive format conversion information as enhancement data in scalable video coding. Experimental results are shown for a wide range of base layer qualities and enhancement bitrates to determine when AFC can improve video scalability. Since the parameters needed for AFC are small compared to residual coding, AFC can provide video scalability at low enhancement layer bitrates that are not possible with residual coding. In addition, AFC can also be used in addition to residual coding to improve video scalability at higher enhancement layer bitrates. Adaptive format conversion has not been studied in detail, but many scalable applications may benefit from it. An example of an application that AFC is well-suited for is the migration path for digital television where AFC can provide immediate video scalability as well as assist future migrations.

A joint temporal-spatial rate control scheme is investigated to optimize temporal and spatial rate-distortion (R-D) performances for full frame-rate video playback, where skipped frames are reconstructed via frame interpolation. The derived optimization problem is too complex to solve by using a straightforward Lagrangian multiplier method due to inter-frame coding dependency (e.g. I/P/B) and the frame-dependent reconstruction. To reduce the complexity, frames are adaptively grouped based on MAD (mean absolute difference) differentials, and iterative greedy search based on the R-D cost is applied to each group of frames to obtain a suboptimal solution. The performance of full frame-rate playback of the ITU-T H.263+/TMN8 codec is evaluated by reconstructing skipped frames with bidirectional motion-compensated interpolation. Experimental results show that the proposed solution has a gain of 0.2 - 1.0 dB over the traditional scheme with a fixed frame skip in average PSNR. Furthermore, the frame-by-frame PSNR variance is also reduced.

Watermarking, traditionally used for copyright protection, is used in a new and exciting way. An efficient wavelet-based watermarking technique embeds audio information into a video signal. Several effective compression techniques are applied to compress the resulting audio/video signal in an embedded fashion. This wavelet-based compression algorithm incorporates bit-plane coding, index coding, and Huffman coding. To demonstrate the potential of this audio embedding and audio/video compression algorithm, we embed an audio signal into a video signal and then compress. Results show that overall compression rates of 15:1 can be achieved. The video signal is reconstructed with a median PSNR of nearly 33 dB. Finally, the audio signal is extracted from the compressed audio/video signal without error.

In this paper, we present a wavelet-based image compression technique, which incorporates some of the human visual system (HVS) characteristics for the wavelet decomposition and bit allocation of subband images. The wavelet coefficients are coded using a new technique, referred to as Block Pattern Coding algorithm. The proposed technique employs a set of local geometric patterns, which preserve the underlying edge geometries in the high frequency signals at very low coding rates. Critical to the success of our approach is the frequent utilization of a special block pattern, which is a uniform pattern of constant intensity to reproduce image blocks of near constant intensity. A performance comparison with JPEG and HVS-based wavelets using VQ is presented for both moderate and heavy compression.

In this paper, we present a pair of new unitary transforms that were derived from the symmetric and antisymmetric orthonormal multiwavelet transform and the discrete cosine transform (DCT). This is motivated by the fact that the current international image compression standard, JPEG, is using DCT, whereas the proposed new standard, JPEG 2000, is using the 9/7 biorthogonal wavelet transforms as the default transform. Yet, recent research has reported that the Lapped transform which uses DCT as building blocks can obtain better performance than the 9/7 biorthogonal wavelet transform. ON the other hand, the relationship between the wavelet transform and DCT is not well known because of completely different paths of evolution. In this paper we explore the connection between the symmetric and antisymmetric orthonormal multiwavelet transform and the DCT. The known multiwavelet transforms today have been limited to the types consisting of only two scaling (dilation) functions and two wavelet functions. Based on the multiwavelet concept a pair of new block transforms, similar to the DCT, can be generated. Through extensive simulations we can show that both of the new unitary transforms perform better than the ordinary DCT. One of the new unitary transforms is preferred not only for its better performance but also fir its nice data-flow architecture. This new unitary transform also leads to a new MWT with four scaling functions and four wavelet functions.

The Cooley-Tukey radix-2 Fast Fourier Transform (FFT) is well known in digital signal processing and has been popularly used in many applications. However, one important function in signal processing is to merge or split of FFT blocks in the Fourier transform domain. The Cooley-Tukey radix-2 decimation-in-frequency FFT algorithm can not be used for this purpose because twiddle factors must be multiplied to the input data before FFT is performed on the resultant. In other words, the existing radix-2 decimation- in-frequency FFT algorithm is not a true radix-2 algorithm. This in turn has prevented it from direct applications to the transform-domain processing, such as merge or split of FFT blocks in the Fourier domain. For real input data one may prefer to use the Fast Hartley Transform (FHT) because it completely deals with real arithmetic calculations. Then the same statements with regard to the radix-2 decimation-in-frequency FFT apply equally well to FHT because the existing FHT algorithms are the real-number equivalence of the complex-number FFT. The true radix-2 Decimation-in-frequency FFT and FHT algorithms presented in this paper have alleviated the above difficulty, and they may provide new techniques for other potential applications.

In this paper, new methods to eliminate boundary artifacts for overlapping trigonometric bases used in image compression are introduced. By applying overlapping cosine-sine-II bases to images instead of non-overlapping cosine-II bases used in the JPEG algorithm, block artifacts can be reduced. In contrast to non-overlapping transforms, an extension of the signal at the signal bounds is necessary. To prevent boundary artifacts in the reconstructed image, the symmetric periodic extension is preferred in image coding. The cosine-II and sine-II basis functions are symmetric, but nevertheless a conventional symmetric periodic extension is not possible, because different basis functions are used in adjacent intervals. In this paper, we derive weighting functions to make the symmetric periodic extension for these bases possible. We show, that compared to the periodic extension, no visible artifacts appear in the reconstructed image if our new approach is used. In addition, we show that the adaptation of the basis functions at the signal boundaries leads to a better quality of the reconstructed signal.

The k-means algorithm is widely used to design image codecs using vector quantization (VQ). In this paper, we focus on an adaptive approach to implement a VQ technique using the online version of k-means algorithm, in which the size of the codebook is adapted continuously to the statistical behavior of the image. Based on the statistical analysis of the feature space, a set of thresholds are designed such that those codewords corresponding to the low-density clusters would be removed from the codebook and hence, resulting in a higher bit-rate efficiency. Applications of this approach would be in telemedicine, where sequences of highly correlated medical images, e.g. consecutive brain slices, are transmitted over a low bit-rate channel. We have applied this algorithm on magnetic resonance (MR) images and the simulation results on a sample sequence are given. The proposed method has been compared to the standard k-means algorithm in terms of PSNR, MSE, and elapsed time to complete the algorithm.

In this paper, we present an approach for classification and indexing of embryonic gene expression pattern images using shape descriptors for retrieval of data in the biological domain. For this purpose, the image is first subjected to a registration process that involves edge fitting and size-standardization. It is followed by segmentation in order to delineate the expression pattern from the cellular background. The moment invariants for the segmented pattern are computed. Image dissimilarity between images is computed based on these moment invariants for each image pair. Area and Centroids of the segmented expression shapes are used to neutralize the invariant behavior of moment invariants during image retrieval. Details of the proposed approach along with analysis of a pilot dataset are presented in this paper.

A layered space-time coding (STC) system is proposed to transmit video signal over wireless channels. An input video sequence is compressed and data-partitioned into layers with different priorities. Then, unequal error protection is incorporated with the space-time block coding to provide different levels of protection to the different layers. At the receiver, a minimum mean square error (MMSE) detector with interference cancellation (IC) is combined with the space-time decoder to reconstruct the signal effectively. We derive the analytic performance for error probability, and conduct the simulation for the transmission of the H.263 video bitstream. It is shown that the unequal error protection enhances the PSNR performance up to 10 dB in moderate signal to noise ratio environment.

Chromosomes are the primary objects studied in cytogenetics. Recent efforts have been devoted to automating the analysis of banded metaphase chromosomes. Feature extraction is the first step to identify a chromosome. Many useful features, such as the length and the number of bands of a chromosome, can be measured along with the chromosome's longitudinal symmetric axis. Therefore, finding this axis is a necessary precursor to making those measurements. In this paper, a new algorithm for finding a symmetric axis of a chromosome is discussed. The author introduced a concept of local symmetric property of an oblong object, and, then, applied this concept to a chromosome to find the symmetric axis after the boundary of the chromosome has been found. The results of the experiments show that the algorithm works well for both straight and bent chromosomes. Since the algorithms is based on the geometric properties of an object rather than its biological properties, it also can be used to find the symmetric axis of any other oblong object.

The subject of this paper is the extraction of houses in very high resolution satellite data. For this purpose, existing segmentation techniques are analyzed as a tool for house detection in IKONOS data. Additionally, a new combination of region and edge based segmentation as well as classification techniques is presented that uses an optimum of the inherent information of the given image data to achieve best possible detection results. Besides spectral and gray value features of the multispectral and panchromatic bands, context and shape information is extracted and incorporated within the classification process.

A new technique for local contrast enhancement using rank-order filters with spatially adaptive neighborhoods is proposed. The technique is based on the unsharp masking operation. However, instead a linear lowpass filtering we use various rank-order smoothing operations. The smoothing is performed over the pixels of spatially adaptive neighborhoods of details to be enhanced and their surrounding backgrounds. Various rank-order filters for local enhancement of small and middle-size details are implemented. Computer simulation results using a real aerial image are provided and discussed.

Radio astronomy interferometric arrays traditionally use Earth rotation aperture image synthesis. Existing radio telescopes consist of dozens antennas separated at hundreds and thousands wavelengths, and these arrays are very sparse comparing to the common radar & communications phased arrays. New projects of superlarge radio telescopes, Square Kilometer Array (SKA), Low Frequency Array (LOFAR), Atacama Large Millimeter Array (ALMA) presume both Earth rotation and snapshot imaging. Optimizing an array configuration is an important stage of the array design. Due to the sparseness of the radio interferometers, the following cost functions might be chosen during optimization process: sidelobe minimization, or, in more specific way, the maximum sidelobe amplitude or the baseline histogram. Genetic algorithm is proposed in this paper for solving the optimization problem. It provides the global maximum of a cost function in a multimodal task and admits easy implementation of different constrains: desirable angular resolution (maximal antenna spacing), sensitivity to extended image features (minimal spacing), topography limitations, etc. Several examples of array configuration optimization using genetic algorithms are given in the paper.

This paper presents a novel image compression scheme based on the perceptual classification of image patterns in the Discrete Hermite Transform (DHT) domain over a roughly hexagonal sampling lattice. The DHT analyzes a signal through a set of binomial filters which approximate the Gaussian derivatives with the advantage that they are computed efficiently. In order to obtain the DHT referred to a rotated coordinate system the set of coefficients of a given order are mapped through a unitary transformation that is locally specified. Such a transformation is based on the generalized binomial functions so that the rotation algorithm is efficient too. This representation allows a perceptual classification, which is achieved by thesholding the approximation errors that are obtained under the hypotheses that the underlying pattern is a constant (0-D), an oriented structure (1-D) or a non-oriented estructure (2-D). The threshold is based on light adaptation and contrast masking properties of the human vision. Then, a compression is obtained by elimination of coefficients that are visually irrelevant.

An automatic accurate recognition scheme for large-scale structures such as ocean currents, eddies and water masses is an essential technology for extracting environmental information and fishing resource. In this paper, new autonomous recognition schemes based on knowledge-base approaches are derived. And a fundamental scheme for enabling processing to be carried out by small-scale computer architectures is developed. These schemes can greatly reduce the amount of data required to describe large-scale structures like ocean eddies and/or ocean currents, and are therefore expected to be very useful for the autonomous on-orbit processing of ocean observation data. Some applications of the schemes to the recognition of moving shapes in noisy images remotely sensed from Earth orbits are also presented with evaluative experimental results.

A novel filtering algorithm applicable to image processing is presented. It was designed using rank-ordered mean (ROM) estimator to remove an outlier and robust local data activity estimators to detect the outliers. The proposed filter effectively remove impulse noise and preserve edge and fine details. The filter possesses good visual quality of the processed simulated images and good quantitative quality in comparison to the standard median filter. Recommendations to obtain best processing results by proper selection of the filter parameters are given. The designed filter is suitable for impulse noise removal in any image processing applications. One can use it at the first stage of image enhancement followed by any detail-preserving techniques such as the Sigma filter at the second stage.

This paper presents a Bayesian method for reconstructing transmission images, which provide attenuation correction factors for emission scans. In order to preserve the edges that bound anatomical regions, which are important especially for areas of non-uniform attenuation, we use the line-process model as a prior. Our prior model provides edge maps containing the anatomical boundary information as well as edge preserved reconstructions. To optimize our nonconvex objective function, we use our previously developed deterministic annealing algorithm, in which the energy function is approximated by a sequence of smooth functions that converges uniformly to the original energy function. To accelerate the convergence speed of our algorithm, we apply the ordered subsets principle to the deterministic annealing algorithm. We also show how the smoothing parameter can be adjusted to account for the effects of using ordered subsets so that the degree of smoothness can be retained for variations of the number of subsets. To validate the quantitative performance of our algorithm, we use the quantitation of bias/variance over noise trials. Our preliminary results indicate that, in some circumstances, our methods have advantages over conventional methods.

There exists invalid region when collected the X-ray image. It is useful to recognize the valid regions and make these regions the only regions to process. Some paper has discussed this question. In additional, we present a new method on recognizing the valid region based on curvature judgment. This method improves the robust of the algorithm. It does well to the X-ray image collection.

Color dropout refers to the process of converting color form documents to black and white by removing the colors that are part of the blank form and maintaining only the information entered in the form. In this paper, no prior knowledge of the form type is assumed. Color dropout is performed by associating darker non-dropout colors with information that is entered in the form and needs to be preserved. The color dropout filter parameters include the color values of the non-dropout colors, e.g. black and blue, the distance metric, e.g. Euclidian, and the tolerances allowed around these colors. Color dropout is accomplished by converting pixels that have color within the tolerance sphere of the non-dropout colors to black and all others to white. This approach lends itself to high-speed hardware implementation with low memory requirements, such as an FPGA platform. Processing may be performed in RGB or a Luminance-Chrominance space, such as YC_bC_r. The color space transformation from RGB to YC_bC_r involves a matrix multiplication and the dropout filter implementation is similar in both cases. Results for color dropout processing in both RGB and YC_bC_r space are presented.

In this paper, we present the real time implementation of the robust RM-estimators with different influence functions such as the cut median, Hampel, Andrews sine, Tukey and Bernoulli functions. The use of these functions in the RM algorithms provides the retention of small-size details, impulsive noise removal and multiplicative noise suppression. They demonstrated better robustness in comparison with the use of the simplest cut function. The optimal values of the parameters of such filters in presence of different noise mixture are determined. The implementation by means of use of DSP TMS320C6701 has demonstrated that the values of time processing in the case of the simplest cut function is less in comparison with another influence functions, but noise suppression is better when the proposed functions were applied.

In this paper, the authors built an experimental test bed to transmit medical image though gigabit Ethernet Optical-fiber communication system, and apply a digital signal processing technology to the medical image. The eye patterns are measured for calculating bit error rate (BER) and signal-to- noise (SNR) of this system. At last, we use the digital signal processing method to extract the medical image feature at the receiver terminal of the transmission system.

The recent MPEG 4 and JPEG 2000 standards address the need for content based coding and manipulation of visual media. The upcoming MPEG 7 standard proposes content descriptors, which succinctly describe the visual content for the purpose of efficient retrieval. This implies that there is an impending need for efficient and effective joint compression and indexing approaches. Several compressed domain indexing techniques have been presented in the recent literature. These are based on the extraction of features from the compression parameters to derive the indices. However, there is little work in the domain of exploring the use of these features to serve the purposes of both compression and indexing. In this paper, we propose a novel technique for joint compression and indexing in the wavelet domain. We not that wavelet based compression is used in JPEG 2000 and (for texture coding) in MPEG 4. In the proposed technique, the wavelet decomposed image is first preprocessed to extract features which are then used for compressing the image as well as for deriving the indices. Extensive simulations have been performed in the JPEG 2000 compressed domain to demonstrate the efficiency of the proposed technique.

In this paper we present a true radix-2 discrete cosine transform (DCT) algorithm in both decimation-in-frequency and decimation-in-time forms. To date, there has been strong interest in developing new processing techniques in the DCT domain for the reason that the DCT is popularly used in the current international standards for image, audio, and video compression. One important function in this respect is to merge or split of DCT blocks in the transform domain. Though many fast DCT algorithms have been existed, they are not suitable for such application. Most of the existing fast DCT algorithms are in radix-2 form, of which the DCT matrix is factorized into two half-sized transform matrices. But these two sub-matrices are not the same, at the best only one is the lower order DCT. In other words, the existing fast DCT algorithms are not true radix-2. This in turn has prevented them from direct applications to the transform-domain processing. The true radix-2 DCT algorithm presented in this paper has alleviated the above difficulty, and it may provide new techniques for other potential applications.

This paper presents an unequal error protection of embedded video bitstreams using three dimensional SPIHT (3-D SPIHT) algorithm and Spatio-Temporal Tree Preserving 3-D SPIHT (STTP-SPIHT) algorithm. We have already proved the efficiency and robustness of the STTP-SPIHT in both of noisy and noiseless channels by modifying the 3-D SPIHT algorithm. We demonstrate that the 3-D SPIHT can also be error resilient against channel bit errors by dividing the embedded video bitstreams, and more error resilient when we divide the STTP-SPIHT bitstreams.

Standard video compression algorithms share one common coding framework based on the hybrid block-based motion-compensated DCT structure. Recently, video compression at low bit-rates has become the focus of research in the signal processing community due to the expanding applications in video conferencing, video over telephone lines, streaming video over the Internet, multimedia and wireless video communications. Unfortunately, in these situations, the notorious blocking artifacts resulting from block DCT and block-based motion estimation/compensation set a severe limit on the achievable bit-rate with acceptable quality. To avoid blocking artifacts and to improve video coding efficiency, this paper presents a novel video compression algorithm based on the lapped transform and overlapping block motion estimation/compensation. Our long-term focus is to develop a complete integrated framework for lapped-transform-based video coding, ranging from theory, design, fast implementations, to practically desirable features for video streaming and delivery over communication networks. In this paper, we are mainly concerned with fundamental theoretical issues. The goals of the proposed video coding technique are to eliminate blocking artifacts and to improve coding efficiency while maintaining a minimal complexity overhead and retaining the flexibility that block-based methods possess. Preliminary coding results confirm the validity of the proposed theory.

It has been well established that state-of-the-art wavelet image coders outperform block transform image coders in the rate-distortion (R-D) sense by a wide margin. Wavelet-based JPEG2000 is emerging as the new high-performance international standard for still image compression. An often asked question is: how much better are wavelets comparing to block transforms in image coding? A notable observation is that each block transform coefficient is highly correlated with its neighbors within the same block as well as its neighbors within the same subband. Current block transform coders such as JPEG suffer from poor context modeling and fail to take full advantage of inter-block correlation in both space and frequency sense. This paper presents a simple, fast and efficient adaptive block transform image coding algorithm based on a combination of pre-filtering, post-filtering, and high-order space-frequency context modeling of block transform coefficients. Despite the simplicity constraints, coding results show that the proposed codec achieves competitive R-D performance comparing to the best wavelet codecs.