• First beam image dataset capturing diverse optical-alignment patterns and parameters • Optical resonator alignment cast as a pairwise beam-pattern regression task • Two-stage model with both feature interaction and refinement for coarse-to-fine alignment • Trained on one device, the model generalizes to another without re-training • Achieves high accuracy with real-time inference on embedded edge hardware Accurate alignment of a laser resonator is essential for upscaling industrial laser manufacturing and precision processing. However, traditional manual or semi-automatic methods depend heavily on operator expertise, and struggle with the interdependence among multiple alignment parameters. To tackle this, we introduce the first real-world image dataset for automatic laser resonator alignment, collected on a laboratory-built resonator setup. It comprises over 6,000 beam profiler images annotated with four key alignment parameters (intracavity iris aperture diameter, output coupler pitch and yaw actuator displacements, and axial position of the output coupler), with over 500,000 paired samples for data‐driven alignment. Given a pair of beam profiler images exhibiting distinct beam patterns under different configurations, the system predicts the control-parameter changes required to realign the resonator. Leveraging this dataset, we propose a novel two-stage deep learning framework for automatic resonator alignment. In Stage 1, a multi-scale CNN augmented with cross-attention and correlation-difference modules, extracts features and outputs an initial coarse prediction of alignment parameters. In Stage 2, a feature-difference map is computed by subtracting the paired feature representations and fed into an iterative refinement module to correct residual misalignments. The final prediction combines coarse and refined estimates, integrating global context with fine-grained corrections for accurate inference. Experiments on our dataset and a different instance of the same physical system from which the CNN was trained suggest superior accuracy and practicality to manual alignment.

This paper establishes a new framework for assessing multimodal statistical causality between cryptocurrency market (cryptomarket) sentiment and cryptocurrency price processes. In order to achieve this, we present an efficient algorithm for multimodal statistical causality analysis based on Multiple-Output Gaussian Processes. Signals from different information sources (modalities) are jointly modelled as a Multiple-Output Gaussian Process, and then using a novel approach to statistical causality based on Gaussian Processes (GPs), we study linear and non-linear causal effects between the different modalities. We demonstrate the effectiveness of our approach in a machine learning application by studying the relationship between cryptocurrency spot price dynamics and sentiment time-series data specific to the crypto sector, which we conjecture influences retail investor behaviour. The investor sentiment is extracted from cryptomarket news data via methods developed in the area of statistical machine learning known as Natural Language Processing (NLP). To capture sentiment, we present a novel framework for text to time-series embedding, which we then use to construct a sentiment index from publicly available news articles. We conduct a statistical analysis of our sentiment statistical index model and compare it to alternative state-of-the-art sentiment models popular in the NLP literature. In regard to the multimodal causality, the investor sentiment is our primary modality of exploration, in addition to price and a blockchain technology-related indicator (hash rate). Analysis shows that our approach is effective in modelling causal structures of variable degree of complexity between heterogeneous data sources and illustrates the impact that certain modelling choices for the different modalities can have on detecting causality. A solid understanding of these factors is necessary to gauge cryptocurrency adoption by retail investors and provide sentiment- and technology-based insights about the cryptocurrency market dynamics.

Estimation of texture similarity is fundamental to many material recognition tasks. This study uses fine-grained human perceptual similarity ground-truth to provide a comprehensive evaluation of 51 texture feature sets. We conduct two types of evaluation and both show that these features do not estimate similarity well when compared against human agreement rates, but that performances are improved when the features are combined using a Random Forest. Using a simple two-stage statistical model we show that few of the features capture long-range aperiodic relationships. We perform two psychophysical experiments which indicate that long-range interactions do provide humans with important cues for estimating texture similarity. This motivates an extension of the study to include Convolutional Neural Networks (CNNs) as they enable arbitrary features of large spatial extent to be learnt. Our conclusions derived from the use of two pre-trained CNNs are: that the large spatial extent exploited by the networks' top convolutional and first fully-connected layers, together with the use of large numbers of filters, confers significant advantage for estimation of perceptual texture similarity.

In this article, we investigate the importance of phase for texture discrimination and similarity estimation tasks. We first use two psychophysical experiments to investigate the relative importance of phase and magnitude spectra for human texture discrimination and similarity estimation. The results show that phase is more important to humans for both tasks. We further examine the ability of 51 computational feature sets to perform these two tasks. In contrast with the psychophysical experiments, it is observed that the magnitude data is more important to these computational feature sets than the phase data. We hypothesise that this inconsistency is due to the difference between the abilities of humans and the computational feature sets to utilise phase data. This motivates us to investigate the application of the 51 feature sets to phase-only images in addition to their use on the original data set. This investigation is extended to exploit Convolutional Neural Network (CNN) features. The results show that our feature fusion scheme improves the average performance of those feature sets for estimating humans' perceptual texture similarity. The superior performance should be attributed to the importance of phase to texture similarity.

Autonomous vehicles and robots are increasingly being deployed to remote, dangerous environments in the energy sector, search and rescue and the military. As a result, there is a need for humans to interact with these robots to monitor their tasks, such as inspecting and repairing offshore wind-turbines. Conversational Agents can improve situation awareness and transparency, while being a hands-free medium to communicate key information quickly and succinctly. As part of our user-centered design of such systems, we conducted an in-depth immersive qualitative study of twelve marine research scientists and engineers, interacting with a prototype Conversational Agent. Our results expose insights into the appropriate content and style for the natural language interaction and, from this study, we derive nine design recommendations to inform future Conversational Agent design for remote autonomous systems.

Automated tools are increasingly being used to generate highly engaging concept maps as an aid to strategic planning and other decision-making tasks. Unless stakeholders can understand the principles of the underlying layout process, however, we have found that they lack confidence and are therefore reluctant to use these maps. In this paper, we present a qualitative study exploring the effect on users' confidence of using data-driven explanation mechanisms, by conducting in-depth scenario-based interviews with ten participants. To provide diversity in stimulus and approach we use two explanation mechanisms based on projection and agglomerative layout methods. The themes exposed in our results indicate that the data-driven explanations improved user confidence in several ways, and that process clarity and layout density also affected users' views of the credibility of the concept maps. We discuss how these factors can increase uptake of automated tools and affect user confidence.

In previous work we developed a method for interior designers to receive image-based feedback about a crowd's emotions when viewing their designs. Although the designers clearly desired a service which provided the new style of feedback, we wanted to find out if an internet crowd would enjoy, and become engaged in, giving emotion feedback this way. In this paper, through a mixed methods study, we expose whether and why internet users enjoy giving emotion feedback using images compared to responding with text. We measured the participants' cognitive styles and found that they correlate with the reported utility and engagement of using images. Those more visual than they are verbal were more engaged by using images to express emotion compared to text. Enlightening qualitative insights reveal, surprisingly, that half of our participants have an appetite for expressing emotions this way, value engagement over clarity, and would use images for emotion feedback in contexts other than design feedback.

Research into creating visualisations that organise ideas into concise concept maps often focuses on implicit mathematical and statistical theories which are built around algorithmic efficacy or visual complexity. Although there are multiple techniques which attempt to mathematically optimise this multi-dimensional problem, it is still unknown how to create concept maps that are immediately understandable to people. In this paper, we present an in-depth qualitative study observing the behaviour and discussing the strategy used by non-expert participants to create, interact, update and communicate a concept map that represents a collection of research ideas. Our results show non-expert individuals create concept maps differently to visualisation algorithms. We found that our participants prioritised narrative, landmarks, abstraction, clarity, and simplicity. Finally, we derive design recommendations from our results which we hope will inspire future algorithms that automatically create more usable and compelling concept maps better suited to the natural behaviours and needs of users.

Imagery and language are often seen as serving different aspects of cognition, with cognitive styles theories proposing that people can be visual or verbal thinkers. Most feedback systems, however, only cater to verbal thinkers. To help rectify this, we have developed a novel method of crowd communication which appeals to those more visual people. Designers can ask a crowd to feedback on their designs using specially constructed image banks to discover the perceptual and emotional theme perceived by possible future customers. A major component of the method is a summarization process in which the crowd's feedback, consisting of a mass of images, is presented to the designer as a digest of representative images. In this paper we describe an experiment showing that these image summaries are as effective as the full image selections at communicating terms. This means that designers can consume the new feedback confident that it represents a fair representation of the total image feedback from the crowd.

Dong et al. examined the ability of 51 computational feature sets to estimate human perceptual texture similarity; however, none performed well for this task. While it is well-known that the human visual system is extremely adept at exploiting longer-range aperiodic (and periodic) "contour" characteristics in images, none of the investigated feature sets exploit higher order statistics (HOS) over larger image regions ( > 19×19 pixels). We, therefore, hypothesise that long-range HOS, in the form of contour data, are useful for perceptual texture similarity estimation. We present the results of a psychophysical experiment that shows that contour data are more important, than local image patches, or global second-order data, to human observers for this task. Inspired by this finding, we propose a set of perceptually motivated image features (PMIF) that encode the long-range HOS computed from spatial and angular distributions of contour segments. We use two perceptual texture similarity estimation tasks to compare PMIF against the 51 feature sets referred to above and four commonly used contour representations. This new feature set is also examined in the context of two additional tasks: sketch-based image retrieval and natural scene recognition. The results show that the proposed feature set performs better, or at least comparably to, all the other feature sets. We attribute this promising performance to the fact that the proposed feature set exploits both short-range and long-range HOS.

Cognitive styles theories suggest that we divide into visual and verbal thinkers. In this paper we describe a method designed to encourage visual communication between designers and their audiences. This new visual feedback method is based on enabling fast intuitive selections by the crowd from image banks when responding to an idea. Visual summarization reduces the massed image choices to a small number of representative images. These summaries are then consumed at a glance by designers receiving the feedback leading to thoughtful reflection on their designs. We report an evaluation using two types of imagery for feedback. Twelve designers took part, receiving visual feedback in response to their designs. In semi-structured interviews they described their interpretation of the feedback, how it inspired them to change their designs and contrasted it with text feedback. Eleven of the twelve designers revealed that they would be enthusiastic users of a service providing this new mode of feedback.

Inspired by studies [4, 23, 40] which compared rankings obtained by search engines and human observers, in this paper we compare texture rankings derived by 51 sets of computational features against perceptual texture rankings obtained from a free-grouping experiment with 30 human observers, using a unify evaluation framework. Experimental results show that the MRSAR [37], VZNEIGHBORHOOD [62], LBPHF [2] and LBPBASIC [3] feature sets perform better than their counterparts. However, none of those feature sets are ideal. The best average G and M measures (measures of ranking accuracy from 0 to 1) [15, 5] obtained are 0.36 and 0.25 respectively. We suggest that this poor performance may be due to the small local neighborhood used to calculate higher-order features which cannot capture the long-range interactions that humans have been shown to exploit [14, 16, 49, 56].

Current interactive media presentations of textiles provide an impoverished communication of their 'textile hand', that is their weight, drape, how they feel to touch. These are complex properties experienced through the visual, tactile, auditory and proprioceptive senses and are currently lost when textile materials are presented in interactive video. This paper offers a new perspective from which the production of multi-touch interactive video representations of the tactile qualities of materials is considered. Through an understanding of hand properties of textiles and how people inherently touch and handle them, we are able to develop methods to animate and bring these properties alive using design methods. Observational studies were conducted, noting gestures consumers used to evaluate textile hand. Replicating the appropriate textile deformations for these gestures in interactive video was explored as a design problem. The resulting digital textile swatches and their interactive behavior were then evaluated for their ability to communicate tactile qualities similar to those of the real textiles.

Texture classification and segmentation have been extensively researched over the last thirty years. Early on the Brodatz album[1] quickly became the de facto standard in which a texture class comprised a set of nonoverlapping sub-images cropped from a single photograph. Later, as the focus shifted to investigating illuminationand pose-invariant algorithms, the CUReT database[3] became popular and the texture class became the set of photographs of a single physical sample captured under a variety of imaging conditions. While extremely successful algorithms have been developed to address classification problems based on these databases, the challenging problem of measuring perceived inter-class texture similarity has rarely been discussed. This paper makes use of a new texture collection[4]. It comprises 334 texture samples, including examples of embossed vinyl, woven wall coverings, carpets, rugs, window blinds, soft fabrics, building materials, product packaging, etc. Additionally, an associated perceptual similarity matrix is provided. This was obtained from a grouping experiment using 30 observers. The similarity scores, S(Ii, I j), for each texture pair were calculated simply by dividing the number of observers that grouped the pair into the same sub-set by the number of observers that had the opportunity do so. A dissimilarity matrix was then defined as dsim(Ii, I j) = 1− S(Ii, I j). Hence dsim(Ii, Ii) = 0 for all images Ii, and dsim(Ii, I j) = 1 if none of the participants grouped images Ii together with I j.

As part of the EPSRC funded 'Digital Sensoria' project a set of digital tools were utilised to better demonstrate the tactile qualities of textiles via the internet. Shoogleit [8], an online utility for the creation of interactive video was one such tool. A Shoogle player for iOS mobile devices was then created (Orzechowski) using an iterative process, during which experiments were carried out to determine if the added interactivity afforded by Shoogleit could more accurately describe textile qualities and thus aid in the creation of a next generation mobile browser for textiles.

Illumination and view dependent texture provide ample information on the appearance of real materials at the cost of enormous data storage requirements. Hence, past research focused mainly on compression and modelling of these data, however, few papers have explicitly addressed the way in which humans perceive these compressed data. We analyzed human gaze information to determine appropriate texture statistics. These statistics were then exploited in a pilot illumination and view direction dependent data compression algorithm. Our results showed that taking into account local texture variance can increase compression of current methods more than twofold, while preserving original realistic appearance and allowing fast data reconstruction.

The use of illumination and view-dependent texture information is recently the best way to capture the appearance of real-world materials accurately. One example is the Bidirectional Texture Function. The main disadvantage of these data is their massive size. In this article, we employ perceptually-based methods to allow more efficient handling of these data. In the first step, we analyse different uniform resampling by means of a psychophysical study with 11 subjects, comparing original data with rendering of a uniformly resampled version over the hemisphere of illumination and view-dependent textural measurements. We have found that down-sampling in view and illumination azimuthal angles is less apparent than in elevation angles and that illumination directions can be down-sampled more than view directions without loss of visual accuracy. In the second step, we analyzed subjects gaze fixation during the experiment. The gaze analysis confirmed results from the experiment and revealed that subjects were fixating at locations aligned with direction of main gradient in rendered stimuli. As this gradient was mostly aligned with illumination gradient, we conclude that subjects were observing materials mainly in direction of illumination gradient. Our results provide interesting insights in human perception of real materials and show promising consequences for development of more efficient compression and rendering algorithms using these kind of massive data.

Bidirectional Texture Functions (BTF) are commonly thought to provide the most realistic perceptual experience of materials from rendered images. The key to providing efficient compression of BTFs is the decision as to how much of the data should be preserved. We use psychophysical experiments to show that this decision depends critically upon the material concerned. Furthermore, we develop a BTF derived metric that enables us to automatically set a material's compression parameters in such a way as to provide users with a predefined perceptual quality. We investigate the correlation of three different BTF metrics with psychophysically derived data. Eight materials were presented to eleven naive observers who were asked to judge the perceived quality of BTF renderings as the amount of preserved data was varied. The metric showing the highest correlation with the thresholds set by the observers was the mean variance of individual BTF images. This metric was then used to automatically determine the material-specific compression parameters used in a vector quantisation scheme. The results were successfully validated in an experiment with six additional materials and eighteen observers. We show that using the psychophysically reduced BTF data significantly improves performance of a PCA-based compression method. On average, we were able to increase the compression ratios, and decrease processing times, by a factor of four without any differences being perceived.

We present a new theory that shows that two images of the same object taken under two different light directions can be made virtually identical by filtering each image by a directional derivative filter. The direction and magnitude of the derivative is generally different for each image, and depends on illumination directions used. The method requires the object surface to be of uniform Lambertian reflectance and a shallow relief, and the light directions used to be sufficiently inclined from the surface macro-normal. For a specific case when the surface consists of spherical patches, the two images can be made identical, for any two light directions used provided that none of them is perpendicular to the viewing axis. We provide some simple experiments which illustrate the validity of this theory. 1

This paper develops new theory for the optimal placement of photometric stereo lighting in the presence of camera noise. We show that for three lights, any triplet of orthogonal light directions minimises the uncertainty in scaled normal computation. The assumptions are that the camera noise is additive and normally distributed, and uncertainty is defined as the expectation of squared distance of scale normal to the ground truth. If the camera noise is of zero mean and variance sigma2the optimal (minimum) uncertainty in the scaled normal is 3sigma2For case of n > 3 lights, we show that the minimum uncertainty is 9sigma2n, and identify sets of light configurations which reach this theoretical minimum

Lambertian photometric stereo with unknown light source parameters is ambiguous. Provided that the object imaged constitutes a surface, the ambiguity is represented by the group of generalised bas-relief (GBR) transformations. We show that this ambiguity is resolved when specular reflection is present in two images taken under two different light source directions. We identify all configurations of the two directional lights which are singular and show that they can easily be tested for. While previous work used optimisation algorithms to apply the constraints implied by the specular reflectance component, we have developed a linear algorithm to achieve this goal. Our theory can be utilised to construct fast algorithms for automatic reconstruction of smooth glossy surfaces.

We present a new texture classification scheme which is invariant to surface-rotation. Many texture classification approaches have been presented in the past that are image-rotation invariant. However, image rotation is not necessarily the same as surface rotation. We have therefore developed a classifier that uses invariants that are derived from surface properties rather than image properties. Previously we developed a scheme that used surface gradient (normal) fields estimated using photometric stereo. In this paper we augment these data with albedo information and also employ an additional feature set: the radial spectrum. We used 30 real textures to test the new classifier. A classification accuracy of 91% was achieved when albedo and gradient 1D polar and radial features were combined. The best performance was also achieved by using 2D albedo and gradient spectra. The classification accuracy is 99%.

The appearance of a rough surface is affected by the direction from which it is lit and texture classifiers should account for this. We propose a classifier that is robust to lighting direction—even when the direction is unknown. An existing model of the dependency of texture features on lighting direction is used to develop a probabilistic model. Given a feature set, the algorithm estimates the most likely illumination direction for each texture class. The likelihoods of each candidate (with their estimated lighting) are compared to classify the sample. The ability of the classifier to identify illuminant direc-tion, and to assign the correct class, was tested on 25 real texture samples. The classifier was able to accurately estimate both the azimuth and the zenith of the light source for most textures and gave a 98 % classification rate. 1

The segmentation of rough surfaces using their reflectance properties is considered. We present a technique to estimate the orientation of surface facets whose reflectance functions are unknown. The reflectance characteristics of each facet are estimated individually allowing this technique to be applied to non-homogeneous surfaces. Non-Lambertian components are attenuated allowing shape estimation with classical photometric stereo. Simulations with rough surfaces rendered with Phong's model indicate that this approach extends the range of reflectance functions to which classical photometric stereo can be applied. The recovered surface derivatives, together with the original intensity images are used to construct reflectance maps. These are used as features for segmentation. A reflectance based classifier is found to be more accurate than an intensity classifier.

In this paper texture analysis techniques are used to segment rough surfaces into regions of homogeneous texture. The performance of three rough surface classifiers was assessed and compared. The classifiers differ in their discrimination as well as their input and computational requirements. Experiments were used to identify the failure modes of the classifiers and to identify which classifier is best suited to a particular task. A series of guidelines for the choice of classifier are presented and justified. 1

The paper assesses the validity of a model, proposed by Kube and Pentland (1988), that relates a rough surface to its image texture. Simulation was used to assess whether a linear approximation is appropriate, and whether the optimal linear filter agrees with the predictions of Kube and Pentland's model. The predictions of the model about the image directionality were also assessed on real images. It was found that a linear model is capable of modeling the imaging process for surfaces of moderate roughness and Lambertian reflectance, and that, subject to a small modification, Kube and Pentland's model accurately predicts the relationship between surface and image spectra.

Many image-rotation invariant texture classification approaches have been presented. However, image rotation is not necessarily the same as surface rotation. This paper proposes a novel scheme that is surface-rotation invariant. It uses magnitude spectra of the partial derivatives of the surface obtained using photometric stereo. Unfortunately the partial derivative operator is directional. It is therefore not suited for direct use as a rotation invariant feature. We present a simple frequency domain method of removing the directional artefacts. Polarograms (polar functions of spectra) are extracted from resulting spectra. Classification is performed by comparing training and classification polarograms over a range of rotations (1° steps over the range 0° to 180°). Thus the system both classifies the test texture and estimates its orientation relative to the relevant training texture. A proof for the removal of directional artefacts from partial derivative spectra is provided. Results obtained using the classification scheme on synthetic and real textures are presented.

Many image-rotation invariant texture classification approaches have been presented previously. This paper proposes a novel surface-rotation invariant scheme. It uses the eigenvalues of a surface's gradient-space distribution as its features. Unlike the partial derivatives, from which they are computed, these eigenvalue features are invariant to surface rotation. First, we show that a simple classifier using a single isotropic feature (grey-level standard deviation) is not invariant to surface rotation. Then a practical surface rotation invariant classifier that uses photometric stereo to estimate surface derivatives is developed. Results for both classifiers are presented.

Rotation of the illuminant source about a subject textured surface can cause catastrophic failure of texture classification schemes. This is due to the variation of texture feature output that can occur when the illuminant direction is varied. This paper uses theory and experiment to show that the outputs of linear texture filters, and their features, are sinusoidal functions of the illuminant tilt angle.

This paper uses theory, simulation, and laboratory experiment, to show that directional illumination, used during the image acquisition process, acts as a directional filter of three dimensional texture. It is shown that the directional characteristics of image texture are not intrinsic to the physical texture being imaged, as they are affected by the direction of the illumination. This result has important implications for texture classification schemes: as many use directional characteristics for discrimination purposes. Variation of illuminant direction is shown to significantly affect a common texture measure: the Laws ' L5E5 operator [1]. 1.

A knowledge-based architecture currently being implemented is reported, that addresses the problems of responsive and time-constrained reasoning within an autonomous vehicle. The architecture is implemented on a multiprocessor system and uses task scheduling to achieve minimum response times to external stimuli. Time-constrained reasoning is achieved by reasoning at different abstraction levels within the multiprocessor architecture, thereby trading off solution quality against solution time.>

A method for calibrating ultrasound and video sensors in the hostile underwater environment is presented. It yields sufficient accuracy to allow transformation of image data in one sensor's image plane to the baseframe, and from there into the other sensor's image plane, a prerequisite for any form of sensor fusion. The feasibility of automatic segmentation of underwater video scenes in combination with sparse ultrasound data has been demonstrated by experiment.>