The Linear Canonical Transform (LCT) describes the effect of any Quadratic Phase System (QPS) on an input optical wavefield. Special cases of the LCT include the fractional Fourier transform (FRT), the Fourier transform (FT) and the Fresnel Transform (FST) describing free space propagation. We have recently published theory for the Discrete Linear Canonical Transform (DLCT), which is to the LCT what the Discrete Fourier Transform (DFT) is to the FT and we have derived the Fast Linear Canonical Transform (FLCT), a NlogN, algorithm for its numerical implementation using an approach similar to that used in deriving the FFT from the DFT. The algorithm is significantly different to the FFT and is based purely on the properties of the LCT and can be used for fast FT, FRT and FST calculations and in the most general case to rapidly calculate the effect of any QPS. In this paper we develop theory making the algorithm recursive for ease of implementation. We derive the FLCT butterfly and graph a flowchart for the recursive algorithm.

The evaluation of the Modulation Transfer Function of JPEG compression provides significant challenges because of its non-linear and non-stationary nature. Previous works have documented the calculation of the First Order Wiener Kernel to estimate the linear portion of the Modulation Transfer Function of JPEG 6b and 2000 and compared that to measurements made using ISO 12233 and traditional edge techniques. The First Order Weiner Kernel was argued as representing the overall pictorial effect of the compression techniques more closely than measurement procedures relying on a small portion of the field of view. This work directly compares the results for JPEG 6b and 2000. Additional work attempts to calculate an 'effective' Point Spread Function based on the estimated MTF for the compression systems. These results are then combined with estimates of RMS noise to produce an approximation for the effective pictorial information capacity (EPIC) of the system. It is shown that there is a good degree of correlation between the actual bit rate used to encode the image and the system EPIC for monochrome images. This correlation breaks down when red, green or blue channels are analyzed individually.

Some results concerning the correction loop of an Adaptive Optics (AO) system for the eye are presented. This is part of a project aiming to study the effects of AO on visual performance, using psychophysical methods. The AO system used in the project is presented. It comprises a Shack-Hartmann sensor, which measures the light deformation after a double-pass in the eye, and two corrective blocks. A Badal optometer coupled with cylindrical lenses is used to remove the main refractive aberrations, while a deformable mirror deals with the remaining aberrations. This system can enable one to carry out psychophysical experiments as a stimulus is viewed by the subject through the same optics. A bimorph mirror has been tested in view of correcting ocular simulations, and recently implemented in the system. The experimental results, consistent with the simulations, yield to a residual root-mean-square wavefront deviation of about 0.06 microns over a 4.8 mm pupil, corresponding to a Strehl ratio of approximately 0.6.

In this paper I present a novel analysis of the signal-to-noise ratio (SNR) for the complex output of a correlator obtained by Fourier filtering an input image with a binary phase-only filter (BPOF). Rather than defining the output variance as being the variance of the real and imaginary components added together, it is defined as being the variance of the complex magnitude. An expression for the complex magnitude variance is obtained using a Taylor series expansion in which higher order terms are discarded. The variance is found to be related to the ratio of eveness to oddness of the BPOF, which in turn is related to the choice of threshold line angle. For a purely even BPOF, the correlator output is real, hence the variance is the same as for the conventional case. As the degree of oddness increases the higher frequency complex amplitudes of the noise are rotated out of phase with the dc component. Consequently, the complex magnitude variance decreases. For a BPOF that is odd the variance is a minimum as the higher frequency noise variations are orthogonal to the dc component. Furthermore, the SNR for the BPOF in this case is an order of magnitude greater than that obtained with a phase-only filter (POF). This last result contradicts the perceived wisdom that the SNR ratio for a BPOF and POF is in the range 4/π²→1. The analysis results are confirmed through simulation.

A key component of any adaptive optics system (AO) for the correction of wavefront aberrations, is the wavefront sensor(WFS). Many systems operate in a mode where a WFS measures the aberrations present in the incoming beam. The required corrections are determined and applied by the wavefront corrector - often a deformable mirror (DM). We wish to develop a wavefront sensor-less correcting system, as derived from the original adaptive optics system of Muller and Buffington. In this experiment we employ a method in which a correcting element with adjustable segments is driven to maximise some function of the image. We employ search algorithms to find the optimal combination of actuator voltages to maximise a certain sharpness metric. The “sharpness” is based on intensity measurements taken with a CCD camera. Results have been achieved using a Nelder-Mead variation of the Simplex algorithm. Preliminary results show that the Simplex algorithm can minimise the aberrations and restore the Airy rings of the imaged point source. Good correction is achieved within 50-100 iterations of the Simplex algorithm. The results are repeatable and so-called “blind” correction of the aberrations is achieved. The correction achieved using various sharpness algorithms laid out by Muller and Buffington are evaluated and presented.

One of the principal successes of computer vision over the past thirty years has been the development of robust techniques for the estimation of the structure of a 3D scene given multiple views of that scene. Holography is an established technique for recording and reconstructing real-world 3D objects. A single hologram encodes multiple perspectives of the scene simultaneously, and hence provides a novel avenue of extension of these traditional computer vision techniques. In this paper, we explore the pontential use of digital holograms in 3D scene reconstruction where particular regions of interest are occluded under particular views. In our experiments we employ both synthetic holograms of artificial scenes, and optically-captured digital holograms of real-world objects. We show that by selecting a particular set of perspectives, determined by the occlusions present in the scene, the original scene can be reconstructed.

A digital watermark is a visible, or preferably invisible, identification code that is permanently embedded in digital media, to prove owner authentication thereby providing a level of document protection. In this paper, we review several approaches for the generation of watermarks using chaotic functions, and in particular, the logistic chaotic function. Using this function, in conjunction with seed management, it is possible to generate chaotic sequences that may be used to create highpass or lowpass digital watermarks. A slight change in the initial conditions will quickly lead to a significant change in the subsequent states of the system, and thus will generate substantially different watermarks. This technique has been shown to offer an added security advantage over the more traditionally generated watermarks created from pseudorandom sequences, in that only the function seed needs to be stored. It also has the advantage that, through examination of the theoretical properties of the function, it is possible to choose seeds that lead to robust, lowpass watermarks. We review various detection techniques including correlation and statistical methods, and present an analysis of the impact of noise present in a model optical detector. The logistic function presented in this paper is ill defined for certain seed values and has not been fully investigated for the purpose of watermark generation. We consider the impact of the theoretical properties of the logistic function for several chaos-based watermark generation techniques, in particular, their highpass and lowpass properties, which when embedded in digital media, are suitable for correlation and statistical based detection methods.

The paper investigates from the perspective of computer science the phase coherence theory (PCT) and phase coherent data-scatter (PCD-S). These techniques were originally developed for the area of optical tensiographic data mining and analysis but have a more general appplication in data mining. These develoments have recently been augmented with the engineering of a software toolkit called TraceMiner. Although the toolkit was originally devised for tensiography it was developed to perform as a generic data mining and analysis application with PCT, PCD-S and a range of other data mining algorithms implemented. To date the toolkit has been utilised in its main application area, tensiography, but has also been applied to UV-visible spectroscopy. This work presents a critical investigation of the general utility of PCT, PCD-S and the toolkit for data mining and analysis. A new application of PCT and the TraceMiner software toolkit to Raman spectroscopy is presented with discussion of the relevant measures and the information provided by the toolkit. This provides more insight into the generic potential of the techniques for data mining. The analysis performed on theoretical Raman data is augmented with a study of experimental Raman data. Raman spectroscopy is used for composition and fault detecton analysis in semiconductor surfaces. Finally, the utility of the PCT technique in comparison with traditional Raman spectroscopy methods is considered together with some more general applications in the field of imaging and machine vision.

A comprehensive characterisation of a complementary metal-oxide semiconductor (CMOS) and digital signal processor (DSP) camera, used typically in machine vision applications, is presented in this paper. The camera consists of a direct read-out CMOS sensor, each pixel giving a direct analogue voltage output related to light intensity, with an analogue-to-digital converter and digital signal processor on the back-end. The camera operates as a stand-alone device using a VGA display; code being pre-programmed to the onboard random access memory of the DSP. High detection rates (kHz) on multiple pixels were achieved, and the relationship between pixel response time and light intensity was quantified. The CMOS sensor, with 1024x1024 pixels randomly addressable in space and time, demonstrated a dynamic logarithmic light intensity sensitivity range of 120dB. Integrating the CMOS camera with a low coherence Michelson interferometer, optical coherence tomography (OCT) axial depth scans have been acquired. The intended application is an imaging device for simple yet functional full-field optical coherence tomography. The advantages of the CMOS sensor are the potential for carrier-based detection, through the very fast pixel response with under-sampling, and the elimination of the electromechanical lateral scanning of conventional OCT by replacing it with electronic pixel scanning.

In this work an image-based photoplethysmography (PPG) system is developed and tested against a conventional finger-based system as commonly used in clinical practise. A PPG is essentially an optical instrument consisting of a near infrared (NIR) source and detector that is capable of tracking blood flow changes in body tissue. When used with a number of wavelengths in the NIR band blood oxygenation changes as well as other blood chemical signatures can be ascertained yielding a very useful device in the clinical realm. Conventionally such a device requires direct contact with the tissue under investigation which eliminates the possibility of its use for applications like wound management where the tissue oxygenation measurement could be extremely useful. To circumnavigate this shortcoming we have developed a CMOS camera-based system, which can successfully extract the PPG signal without contact with the tissue under investigation. A comparison of our results with conventional techniques has yielded excellent results.

During the rehabilitation process of the post-stroke patients is conducted, their movements need to be localized and learned so that incorrect movement can be instantly modified or tuned. Therefore, tracking these movement becomes vital and necessary for the rehabilitative course. During human movement tracking, the position estimation of human movement is very important. In this paper, the character of the human movement system is first analyzed. Next, camera and inertial sensor are used to respectively measure the position of human movement, and the Kalman filter algorithm is proposed to fuse the two measurement to get a optimization estimation of the position. In the end, the performance of the method is analyzed.

Non-destructive evaluation is widely used in the manufacturing industry for the detection and characterisation of defects. Typical techniques include visual, magnetic particle, fluorescent dye penetrant, ultrasonic, and eddy current inspection. This paper presents a multi-agent approach to combining image data such as these for quality control. The use of distributed agents allows the speed benefits of parallel processing to be realised, facilitating increased levels of detection through the use of high resolution images. The integration of multi-sensor devices and the fusion of their multi-modal outputs has the potential to provide an increased level of certainty in defect detection and identification. It may also allow the detection and identification of defects that cannot be detected by an individual sensor. This would reduce uncertainty and provide a more complete picture of aesthetic and structural integrity than is possible from a single data source. A blackboard architecture, DARBS (Distributed Algorithmic and Rule-based Blackboard System), has been used to manage the processing and interpretation of image data. Rules and image processing routines are allocated to intelligent agents that communicate with each other via the blackboard, where the current understanding of the problem evolves. Specialist agents register image segments into a common coordinate system. An intensity-based algorithm eliminates the landmark extraction that would be required by feature-based registration techniques. Once registered, pixel-level data fusion is utilised so that both complementary and redundant data can be exploited. The modular nature of the blackboard architecture allows additional sensor data to be processed by the addition or removal of specialised agents.

Electroplating has a long tradition in manufacturing which has advanced from a decorative finishing process to a process that applies precious metals in specific areas with the net effect of changing the properties of the electroplated component. The challenges facing electroplating as a manufacturing process are both cost of manufacture and the assurance of quality. If one is to broadly examine the operation of a reel-to-reel electroplating line, a huge dependence on output quality is placed on the human operation aspect. In this paper focus was placed on removing some of the over dependence placed on the operator. Particularly in relation to the visual inspection process by developing a more consistent detection system with increased reliability and repeatability. This paper presents an innovative approach where in-line non-destructive techniques for evaluating metal deposit quality are developed through color vision. To establish the capabilities of color vision in electroplating a series of empirical tests within the production environment gave a good insight into the critical variables. To optimize the vision system, image control techniques were applied these included variation of illumination sources and intensity, part position and camera location. Finally to achieve full traceability of defects recorded, dedicated software was designed and developed to provide both feedback to the operator and database storage of results. Color vision proves to be a viable detection technique for electroplating but each program is part specific and variations in the formed shape of the stamped part can hinder the success of the system. The significance of this vision application cannot be underestimated when one appreciates electroplating defects are predominately visual and this approach works to prevent visual defects being used in assembled connectors.

This paper proposes an integrated framework for the high level design of high performance signal processing algorithms' implementations on FPGAs. The framework emerged from a constant need to rapidly implement increasingly complicated algorithms on FPGAs while maintaining the high performance needed in many real time digital signal processing applications. This is particularly important for application developers who often rely on iterative and interactive development methodologies. The central idea behind the proposed framework is to dynamically integrate high performance structural hardware description languages with higher level hardware languages in other to help satisfy the dual requirement of high level design and high performance implementation. The paper illustrates this by integrating two environments: Celoxica's Handel-C language, and HIDE, a structural hardware environment developed at the Queen's University of Belfast. On the one hand, Handel-C has been proven to be very useful in the rapid design and prototyping of FPGA circuits, especially control intensive ones. On the other hand, HIDE, has been used extensively, and successfully, in the generation of highly optimised parameterisable FPGA cores. In this paper, this is illustrated in the construction of a scalable and fully parameterisable core for image algebra's five core neighbourhood operations, where fully floorplanned efficient FPGA configurations, in the form of EDIF netlists, are generated automatically for instances of the core. In the proposed combined framework, highly optimised data paths are invoked dynamically from within Handel-C, and are synthesized using HIDE. Although the idea might seem simple prima facie, it could have serious implications on the design of future generations of hardware description languages.

This paper presents a portable mobile stereo vision system designed for the assessment of road signage and delineation (lines and reflective pavement markers or "cat's eyes"). This novel system allows both geometric and photometric measurements to be made on objects in a scene. Global Positioning System technology provides important location data for any measurements made. Using the system it has been shown that road signs can be classfied by nature of their reflectivity. This is achieved by examining the changes in the reflected light intensity with changes in range (facilitated by stereo vision). Signs assessed include those made from retro-reflective materials, those made from diffuse reflective materials and those made from diffuse reflective matrials with local illumination. Field-testing results demonstrate the systems ability to classify objects in the scene based on their reflective properties. The paper includes a discussion of a physical model that supports the experimental data.

This paper details a procedure for classifying facial expressions. This is a growing and relatively new type of problem within computer vision. One of the fundamental problems when classifying facial expressions in previous approaches is the lack of a consistent method of measuring expression. This paper solves this problem by the computation of the Facial Expression Shape Model (FESM). This statistical model of facial expression is based on an anatomical analysis of facial expression called the Facial Action Coding System (FACS). We use the term Action Unit (AU) to describe a movement of one or more muscles of the face and all expressions can be described using the AU's described by FACS. The shape model is calculated by marking the face with 122 landmark points. We use Principal Component Analysis (PCA) to analyse how the landmark points move with respect to each other and to lower the dimensionality of the problem. Using the FESM in conjunction with Support Vector Machines (SVM) we classify facial expressions. SVMs are a powerful machine learning technique based on optimisation theory. This project is largely concerned with statistical models, machine learning techniques and psychological tools used in the classification of facial expression. This holistic approach to expression classification provides a means for a level of interaction with a computer that is a significant step forward in human-computer interaction.

We designed and built an experimental setup for subjective (manually controlled) correction of up to 12 orthogonal terms (excluding tilts and defocus) of the aberration of the human eye. In our experiments, the subject was looking through an adaptive optical system at the USAF resolution chart. By using the arrow keys of the computer keyboard, the subject was able to control the amplitudes of up to 12 orthogonal aberration terms, introduced by a deformable mirror, optically conjugated to the pupil of the eye. Preliminary statistical analysis on a group of 12 participants with 6 correction attempts per participant, demonstrated satisfactory correlation of aberration coefficients obtained by the same person in different correction attempts. The majority of the participants were able to improve the visual acuity by subjective optimization of the figure of the deformable mirror.

As whole body MRI (WB-MRI) gains currency, the data this class of technique generates presents new challenges for the imaging community. One acquisition protocol currently being applied with considerable success entails imaging the subject in a number of successive coronal sections, resulting in a high resolution, gap free, full body acquisition. However this technique often results in considerable greylevel offsets between adjacent coronal sections. To make the images suitable for the application of automated image analysis procedures these discontinuities in the grey data must be alleviated. We examine the issues related to this problem, and present a solution based on histogram rescaling, which is designed to correct for the non-uniformities while preserving the integrity of the data histogram so that it can be used robustly in later processing steps. The final datasets reconstructed from the resampled coronal sections exhibit superior greyscale homogeneity, visually and in statistical measures, and the image segmentation results achieved using this corrected data are consistently more robust and more accurate than those arrived at using the original raw data. The approach has been tested and successfully validated on a database of sixty two WB-MRI datasets.

Since the discovery of X rays radiotherapy has had the same aim - to deliver a precisely measured dose of radiation to a defined tumour volume with minimal damage to surrounding healthy tissue. Recent developments in radiotherapy such as intensity modulated radiotherapy (IMRT) can generate complex shapes of dose distributions. Until recently it has not been possible to verify that the delivered dose matches the planned dose. However, one often wants to know the real three-dimensional dose distribution. Three-dimensional radiation dosimeters have been developed since the early 1980s. Most chemical formulations involve a radiosensitive species immobilised in space by gelling agent. Magnetic Resonance Imaging (MRI) and optical techniques have been the most successful gel scanning techniques so far. Optical techniques rely on gels changing colour once irradiated. Parallel beam optical tomography has been developed at the University of Surrey since the late 1990s. The apparatus involves light emitting diode light source collimated to a wide (12cm) parallel beam. The beam is attenuated or scattered (depending on the chemical formulation) as it passes through the gel. Focusing optics projects the beam onto a CCD chip. The dosimeter sits on a rotation stage. The tomography scan involves continuously rotating the dosimeter and taking CCD images. Once the dosimeter has been rotated over 180 degrees the images are processed by filtered back projection. The work presented discusses the optics of the apparatus in more detail.

Image compression aims to reduce the number of information size required to represent an image by reducing redundancies and non pertinent information so as to reduce the space in memory as well as in the medium of transmission, which will be able to plan to treat processes in real time. The compression of the data is currently the very active research object in the world. The purpose of this research is to improve quality of these compressed images and to reduce the time necessary for this operation, by developing a new techniques of compressions, like those based on optics for example. Indeed, it has been known for a long time that the coherent optics, thanks to treating parallel inherent and characteristic a convergent lens to carry out the Fourier Transform, could reduce the treatment times. In order to use the optical properties, we have developed an all-optical compression architecture based on the JPEG standards. This method is based on Fourier Transform which will be the core principle in the realisation of the DCT. After validation of our method by simulations, we will present, in this communication, its optical implementation. An all optical set-up with a convergent lens for the Fourier transform and an EASLM modulator for the opto-electronic interface have been used.

FPGA technology enjoys both the high performance of a dedicated hardware solution and the flexibility of software that is offered by its inherent reprogrammability feature. Image Processing is one application area that can benefit greatly from FPGAs performance and flexibility. This paper presents the design and implementation of a high-level reconfigurable image coprocessor on FPGAs using the Handel-C hardware language. The latter allows non hardware specialists to program FPGAs at a high level using a C-like syntax, albeit with hardware architectures in mind. It hence allows for rapid development of FPGA applications. This is illustrated in this paper in the case of an image coprocessor whose instruction set is based on the operators of Image Algebra. Central to this instruction set are the five core neighbourhood operations of Image Algebra: Convolution, Additive Maximum, Additive Minimum, Multiplicative Maximum and Multiplicative Minimum. These are parameterised in terms of the neighbourhood operation’s window coefficients, window size and image size. Handel-C language was used to design the Image Coprocessor with a fully tested prototype on Celoxica Virtex-E based RC1000-PP PCI board. The paper presents an overview of the approach used to generate FPGA architectures dynamically for the Image Coprocessor using Handel-C, as well as a sample of implementation results.

Mutual information-based image registration has been verified to be quite effective in many clinical applications. However, when calculating the mutual information between two working images, we need to estimate the grey values of the transformed image by interpolation on the reference image, which introduces regular artefacts in the registration function. In this paper, we analyse the underling mechanism of the artefacts, and present a new statistical interpolation, which will not introduce new intensities. In addition, it also breaks the conformity of the interpolation points, which is considered as a major contributing factor to the artefacts in commonly used interpolations. These characteristics make the registration function much smoother, enabling easier convergence to a global extreme. Experimental results on clinical images verify these advantages.

The optical images compression was the subject of several research in the last decade because the coherent light, with its speed and the property to carry out a two-dimensional Fourier transformation instantaneously with coherent optics on the one hand (thanks to a convergent lens) and new technologies on the electro-optical interfaces on the other hand (modulating space of light SLM), makes it possible to quickly compress generally colored images. It gives us a real asset because the compression time is decreased and the field of the images to be compressed is increased. In this article we present an all optical method of colored images compression based on JPEG standard. This architecture will be validated by simulation and an optical setup of implementation will be proposed.

Abundance variations of carbon and nitrogen in globular star clusters provide astronomers with a means to determine a cluster's evolutionary past. Moreover, these clusters are so ancient (~13 billion years) and so well preserved that they provide an ideal diagnostic for the overall chemical history of the Milky Way Galaxy. Traditionally, spectroscopy is the preferred method to perform investigations into such theories. However, it is not without its drawbacks: spectroscopy can normally only be obtained star by star, and both large telescopes and a great deal of time is required to carry out research in this manner. As globular clusters are known to contain up to a million stars, studying each star individually would take too much time to return a true representative sample of the cluster stars. So, we opt instead for a spectrophotometric technique and a statistical approach to infer a cluster's composition variations. This has required the design and use of new custom narrow-band filters centered on the CH and CN molecular absorption bands or their adjacent continua. Two Galactic clusters (M71 & M92) with contrasting characteristics have been chosen for this study. In order to process this data a header-driven (i.e. automated) astronomical data-processing pipeline was developed for use with a family of CCD instruments known as the FOSCs. The advent of CCD detectors has allowed astronomers to generate large quantities of raw data on a nightly basis, but processing of this amount of data is extremely time and resource intensive. In our case the majority of our cluster data has been obtained using the BFOSC instrument on the 1.52m Cassini Telescope at Loiano, Italy. However, as there are a number of these FOSC instruments throughout the world, our pipeline can be easily adapted to suit any of them. The pipeline has been tested using various types of data ranging from brown dwarf stars to globular cluster images, with each new dataset providing us with new problems/bugs to solve and overcome. The pipeline performs various tasks such as data reduction including image de-fringing, image registration and photometry, with final products consisting of RGB colour images and colour magnitude diagrams (CMD).

A robust least squares motion detection algorithm was evaluated with respect to target size, contrast and sensor noise. In addition, the importance of robust motion estimation was also investigated. The test sequences used for the evaluation were generated synthetically to simulate a forward looking airborne sensor moving with translation parallel to a flat background scene with an inserted target moving orthogonal to the camera motion. For each evaluation parameter, test sequences were generated and from the processed imagery the algorithm performance measured by calculating a receiver-operating-characteristic curve. Analysis of the results revealed that the presence of small amounts of noise results in poor performance. Other conclusions are that the algorithm performs extremely well following noise reduction, and that target contrast has little effect on performance. The system was also tested on several real sequences for which excellent segmentation was obtained. Finally, it was found that for small targets and a downward looking sensor, the performance of the basic least squares was only slightly inferior to the robust version. For larger targets and a forward looking sensor the robust version performed significantly better.

A few fundamental symbols are common to many mesoamerican cultures. However the shape of each symbol greatly varies depending on the considered culture and period. In this paper, we present a technique to recognize one of these symbols in spite of its shape variability. We base the recognition on a small set of symmetry and morphology rules. We describe the rules and their computational implementation, as well as practical recognition results.

This paper presents a development platform for real-time image processing based on the ADSP-BF533 Blackfin processor and the MicroC/OS-II real-time operating system (RTOS). MicroC/OS-II is a completely portable, ROMable, pre-emptive, real-time kernel. The Blackfin Digital Signal Processors (DSPs), incorporating the Analog Devices/Intel Micro Signal Architecture (MSA), are a broad family of 16-bit fixed-point products with a dual Multiply Accumulate (MAC) core. In addition, they have a rich instruction set with variable instruction length and both DSP and MCU functionality thus making them ideal for media based applications. Using the MicroC/OS-II for task scheduling and management, the proposed system can capture and process raw RGB data from any standard 8-bit greyscale image sensor in soft real-time and then display the processed result using a simple PC graphical user interface (GUI). Additionally, the GUI allows configuration of the image capture rate and the system and core DSP clock rates thereby allowing connectivity to a selection of image sensors and memory devices. The GUI also allows selection from a set of image processing algorithms based in the embedded operating system.

Phase Coherent Data-scatter (PCD-S) was originally developed for the area of tensiographic data mining and analysis. This development has been augmented with the engineering of a software toolkit called TraceMiner, which integrates this technique with additional data mining and statistical tools for general use. This paper presents, for the first time, a theoretical treatment of data-scatter as a generic data mining tool, cognisant of the data set descriptions, data transformations, measurands and data model visualisations possible with data-scatter. Data-diffraction resulting from data scatter is also presented here for the first time. The use of the two approaches in a Hough technique to analyse the resulting data-diffraction patterns is discussed briefly in the context of applications of this new data scatter approach.

We propose an algorithm for phase retrieval from three interferograms which differ only by an arbitrary unknown tilt terms in the phase. The method is illustrated by examples.

Multispectral imaging produces large amounts of data which extend processing, transmission and storage systems to their upper limits. Although there are several interface standards specific to image data acquisition, such as CameraLink, it is Firewire which provides a high-speed data bus, integrated control capability, without loss of flexibility, and which is commonly available as a low cost solution. The class of multispectral imaging requires a different treatment of the processing principals than standard imaging. The same spatial region is captured multiple times using different optical wavelengths. This technique finds application in such diverse areas as coastal monitoring, fruit sorting and automated agriculture. Modifications and additional features to the camera operating and configuration parameters are therefore required which are not generally present with conventional imaging sensors. This paper describes extensions to the IIDC Digital Camera (D-Cam) specification in the development of a Firewire technology platform for transmitting the data structures described and for providing real-time, online control of spectral information acquisition. Additionally, it describes how a set of registers in the sub-unit architecture of the Firewire protocol is augmented to accommodate the demands of a multispectral system. The extensions are specification conformant and do not alter underlining compliance with the base standard. The paper also describes the implementation of the extended D-Cam in the Firewire subsystem of a smart multispectral camera used in commercial applications.

In this paper we discuss a microprocessor controlled illumination system that covers the visible spectrum. The system uses multiple discrete wavelength LEDs. Uniform intensity illumination is achieved by controlling LED currents with an array of voltage controlled current sources. A highly efficient polycarbonate holographic diffuser is used to improve the uniformity of the distribution. Current control is also used to maintain a spatially and spectrally homogenous illumination. Image analysis techniques are then used to parameterise the recorded scene and provide a figure of merit for the quality of the illumination. Several functions such as contrast, low/high saturation, average brightness and histogram entropy are applied to extract objective quality judgements and to compute an overall quality function. This process is then iterated so as to maximise the image quality against embedded quality function benchmarks.

A fully automated volumetric image segmentation algorithm is proposed which uses Bayesian inference to assess the appropriate number of image segments. The segmentation is performed exclusively within the wavelet domain, after the application of the redundant a trous wavelet transform employing four decomposition levels. This type of analysis allows for the evaluation of spatial relationships between objects in an image at multiple scales, exploiting the image characteristics matched to a particular scale. These could possibly go undetected in other analysis techniques. The Bayes Information Criterion (BIC) is calculated for a range of segment numbers with a relative maximum determining optimal segment number selection. The fundamental idea of the BIC is to approximate the integrated likelihood in the Bayes factor and then ignore terms which do not increase quickly with N, where N is the cardinality of the data. Gaussian Mixture Modelling (GMM) is then applied to an individual mid-level wavelet scale to achieve a baseline scene estimate considering only voxel intensities. This estimate is then refined using a series of wavelet scales in a multiband manner to reflect spatial and multiresolution correlations within the image, by means of a Markov Random Field Model (MRFM). This approach delivers promising results for a number of volumetric brain MR and PET images, with inherent image features being identified. Results achieved largely correspond with those obtained by researchers in biomedical imaging utilising manually defined parameters for image modelling.