The role of context in image annotation and recommendation

With the rise of smart phones, lifelogging devices (e.g. Google Glass) and popularity of image sharing websites (e.g. Flickr), users are capturing and sharing every aspect of their life online producing a wealth of visual content. Of these uploaded images, the majority are poorly annotated or exist in complete semantic isolation making the process of building retrieval systems difficult as one must firstly understand the meaning of an image in order to retrieve it. To alleviate this problem, many image sharing websites offer manual annotation tools which allow the user to “tag” their photos, however, these techniques are laborious and as a result have been poorly adopted; Sigurbjörnsson and van Zwol (2008) showed that 64% of images uploaded to Flickr are annotated with < 4 tags. Due to this, an entire body of research has focused on the automatic annotation of images (Hanbury, 2008; Smeulders et al., 2000; Zhang et al., 2012a) where one attempts to bridge the semantic gap between an image’s appearance and meaning e.g. the objects present. Despite two decades of research the semantic gap still largely exists and as a result automatic annotation models often offer unsatisfactory performance for industrial implementation. Further, these techniques can only annotate what they see, thus ignoring the “bigger picture” surrounding an image (e.g. its location, the event, the people present etc). Much work has therefore focused on building photo tag recommendation (PTR) methods which aid the user in the annotation process by suggesting tags related to those already present. These works have mainly focused on computing relationships between tags based on historical images e.g. that NY and timessquare co-exist in many images and are therefore highly correlated. However, tags are inherently noisy, sparse and ill-defined often resulting in poor PTR accuracy e.g. does NY refer to New York or New Year? This thesis proposes the exploitation of an image’s context which, unlike textual evidences, is always present, in order to alleviate this ambiguity in the tag recommendation process. Specifically we exploit the “what, who, where, when and how” of the image capture process in order to complement textual evidences in various photo tag recommendation and retrieval scenarios. In part II, we combine text, content-based (e.g. # of faces present) and contextual (e.g. day-of-the-week taken) signals for tag recommendation purposes, achieving up to a 75% improvement to precision@5 in comparison to a text-only TF-IDF baseline. We then consider external knowledge sources (i.e. Wikipedia & Twitter) as an alternative to (slower moving) Flickr in order to build recommendation models on, showing that similar accuracy could be achieved on these faster moving, yet entirely textual, datasets. In part II, we also highlight the merits of diversifying tag recommendation lists before discussing at length various problems with existing automatic image annotation and photo tag recommendation evaluation collections. In part III, we propose three new image retrieval scenarios, namely “visual event summarisation”, “image popularity prediction” and “lifelog summarisation”. In the first scenario, we attempt to produce a rank of relevant and diverse images for various news events by (i) removing irrelevant images such memes and visual duplicates (ii) before semantically clustering images based on the tweets in which they were originally posted. Using this approach, we were able to achieve over 50% precision for images in the top 5 ranks. In the second retrieval scenario, we show that by combining contextual and content-based features from images, we are able to predict if it will become “popular” (or not) with 74% accuracy, using an SVM classifier. Finally, in chapter 9 we employ blur detection and perceptual-hash clustering in order to remove noisy images from lifelogs, before combining visual and geo-temporal signals in order to capture a user’s “key moments” within their day. We believe that the results of this thesis show an important step towards building effective image retrieval models when there lacks sufficient textual content (i.e. a cold start).

Drivers of technology substitution : successive generations of high tech products

Principal Topic High technology consumer products such as notebooks, digital cameras and DVD players are not introduced into a vacuum. Consumer experience with related earlier generation technologies, such as PCs, film cameras and VCRs, and the installed base of these products strongly impacts the market diffusion of the new generation products. Yet technology substitution has received only sparse attention in the diffusion of innovation literature. Research for consumer durables has been dominated by studies of (first purchase) adoption (c.f. Bass 1969) which do not explicitly consider the presence of an existing product/technology. More recently, considerable attention has also been given to replacement purchases (c.f. Kamakura and Balasubramanian 1987). Only a handful of papers explicitly deal with the diffusion of technology/product substitutes (e.g. Norton and Bass, 1987: Bass and Bass, 2004). They propose diffusion-type aggregate-level sales models that are used to forecast the overall sales for successive generations. Lacking household data, these aggregate models are unable to give insights into the decisions by individual households - whether to adopt generation II, and if so, when and why. This paper makes two contributions. It is the first large-scale empirical study that collects household data for successive generations of technologies in an effort to understand the drivers of adoption. Second, in comparision to traditional analysis that evaluates technology substitution as an ''adoption of innovation'' type process, we propose that from a consumer's perspective, technology substitution combines elements of both adoption (adopting the new generation technology) and replacement (replacing the generation I product with generation II). Based on this proposition, we develop and test a number of hypotheses. Methodology/Key Propositions In some cases, successive generations are clear ''substitutes'' for the earlier generation, in that they have almost identical functionality. For example, successive generations of PCs Pentium I to II to III or flat screen TV substituting for colour TV. More commonly, however, the new technology (generation II) is a ''partial substitute'' for existing technology (generation I). For example, digital cameras substitute for film-based cameras in the sense that they perform the same core function of taking photographs. They have some additional attributes of easier copying and sharing of images. However, the attribute of image quality is inferior. In cases of partial substitution, some consumers will purchase generation II products as substitutes for their generation I product, while other consumers will purchase generation II products as additional products to be used as well as their generation I product. We propose that substitute generation II purchases combine elements of both adoption and replacement, but additional generation II purchases are solely adoption-driven process. Extensive research on innovation adoption has consistently shown consumer innovativeness is the most important consumer characteristic that drives adoption timing (Goldsmith et al. 1995; Gielens and Steenkamp 2007). Hence, we expect consumer innovativeness also to influence both additional and substitute generation II purchases. Hypothesis 1a) More innovative households will make additional generation II purchases earlier. 1 b) More innovative households will make substitute generation II purchases earlier. 1 c) Consumer innovativeness will have a stronger impact on additional generation II purchases than on substitute generation II purchases. As outlined above, substitute generation II purchases act, in part like a replacement purchase for the generation I product. Prior research (Bayus 1991; Grewal et al 2004) identified product age as the most dominant factor influencing replacements. Hence, we hypothesise that: Hypothesis 2: Households with older generation I products will make substitute generation II purchases earlier. Our survey of 8,077 households investigates their adoption of two new generation products: notebooks as a technology change to PCs, and DVD players as a technology shift from VCRs. We employ Cox hazard modelling to study factors influencing the timing of a household's adoption of generation II products. We determine whether this is an additional or substitute purchase by asking whether the generation I product is still used. A separate hazard model is conducted for additional and substitute purchases. Consumer Innovativeness is measured as domain innovativeness adapted from the scales of Goldsmith and Hofacker (1991) and Flynn et al. (1996). The age of the generation I product is calculated based on the most recent household purchase of that product. Control variables include age, size and income of household, and age and education of primary decision-maker. Results and Implications Our preliminary results confirm both our hypotheses. Consumer innovativeness has a strong influence on both additional purchases (exp = 1.11) and substitute purchases (exp = 1.09). Exp is interpreted as the increased probability of purchase for an increase of 1.0 on a 7-point innovativeness scale. Also consistent with our hypotheses, the age of the generation I product has a dramatic influence for substitute purchases of VCR/DVD (exp = 2.92) and a strong influence for PCs/notebooks (exp = 1.30). Exp is interpreted as the increased probability of purchase for an increase of 10 years in the age of the generation I product. Yet, also as hypothesised, there was no influence on additional purchases. The results lead to two key implications. First, there is a clear distinction between additional and substitute purchases of generation II products, each with different drivers. Treating these as a single process will mask the true drivers of adoption. For substitute purchases, product age is a key driver. Hence, implications for marketers of high technology products can utilise data on generation I product age (e.g. from warranty or loyalty programs) to target customers who are more likely to make a purchase.

Motivation, development and validation of a new spectral greenness index : a spectral dimension related to foliage projective cover

A method is presented for the development of a regional Landsat-5 Thematic Mapper (TM) and Landsat-7 Enhanced Thematic Mapper plus (ETM+) spectral greenness index, coherent with a six-dimensional index set, based on a single ETM+ spectral image of a reference landscape. The first three indices of the set are determined by a polar transformation of the first three principal components of the reference image and relate to scene brightness, percent foliage projective cover (FPC) and water related features. The remaining three principal components, of diminishing significance with respect to the reference image, complete the set. The reference landscape, a 2200 km2 area containing a mix of cattle pasture, native woodland and forest, is located near Injune in South East Queensland, Australia. The indices developed from the reference image were tested using TM spectral images from 19 regionally dispersed areas in Queensland, representative of dissimilar landscapes containing woody vegetation ranging from tall closed forest to low open woodland. Examples of image transformations and two-dimensional feature space plots are used to demonstrate image interpretations related to the first three indices. Coherent, sensible, interpretations of landscape features in images composed of the first three indices can be made in terms of brightness (red), foliage cover (green) and water (blue). A limited comparison is made with similar existing indices. The proposed greenness index was found to be very strongly related to FPC and insensitive to smoke. A novel Bayesian, bounded space, modelling method, was used to validate the greenness index as a good predictor of FPC. Airborne LiDAR (Light Detection and Ranging) estimates of FPC along transects of the 19 sites provided the training and validation data. Other spectral indices from the set were found to be useful as model covariates that could improve FPC predictions. They act to adjust the greenness/FPC relationship to suit different spectral backgrounds. The inclusion of an external meteorological covariate showed that further improvements to regional-scale predictions of FPC could be gained over those based on spectral indices alone.

Spherical image-based visual servo and structure estimation

This paper presents a formulation of image-based visual servoing (IBVS) for a spherical camera where coordinates are parameterized in terms of colatitude and longitude: IBVSSph. The image Jacobian is derived and simulation results are presented for canonical rotational, translational as well as general motion. Problems with large rotations that affect the planar perspective form of IBVS are not present on the sphere, whereas the desirable robustness properties of IBVS are shown to be retained. We also describe a structure from motion (SfM) system based on camera-centric spherical coordinates and show how a recursive estimator can be used to recover structure. The spherical formulations for IBVS and SfM are particularly suitable for platforms, such as aerial and underwater robots, that move in SE(3).

Dynamic Image-based visual servo control using centroid and optic flow features

This paper considers the question of designing a fully image-based visual servo control for a class of dynamic systems. The work is motivated by the ongoing development of image-based visual servo control of small aerial robotic vehicles. The kinematics and dynamics of a rigid-body dynamical system (such as a vehicle airframe) maneuvering over a flat target plane with observable features are expressed in terms of an unnormalized spherical centroid and an optic flow measurement. The image-plane dynamics with respect to force input are dependent on the height of the camera above the target plane. This dependence is compensated by introducing virtual height dynamics and adaptive estimation in the proposed control. A fully nonlinear adaptive control design is provided that ensures asymptotic stability of the closed-loop system for all feasible initial conditions. The choice of control gains is based on an analysis of the asymptotic dynamics of the system. Results from a realistic simulation are presented that demonstrate the performance of the closed-loop system. To the author's knowledge, this paper documents the first time that an image-based visual servo control has been proposed for a dynamic system using vision measurement for both position and velocity.

A new approach for finding an appropriate combination of texture parameters for classification

This paper suggests an approach for finding an appropriate combination of various parameters for extracting texture features (e.g. choice of spectral band for extracting texture feature, size of the moving window, quantization level of the image, and choice of texture feature etc.) to be used in the classification process. Gray level co-occurrence matrix (GLCM) method has been used for extracting texture from remotely sensed satellite image. Results of the classification of an Indian urban environment using spatial property (texture), derived from spectral and multi-resolution wavelet decomposed images have also been reported. A multivariate data analysis technique called ‘conjoint analysis’ has been used in the study to analyze the relative importance of these parameters. Results indicate that the choice of texture feature and window size have higher relative importance in the classification process than quantization level or the choice of image band for extracting texture feature. In case of texture features derived using wavelet decomposed image, the parameter ‘decomposition level’ has almost equal relative importance as the size of moving window and the decomposition of images up to level one is sufficient and there is no need to go for further decomposition. It was also observed that the classification incorporating texture features improves the overall classification accuracy in a statistically significant manner in comparison to pure spectral classification.

Ontology-based image annotation

With regard to the long-standing problem of the semantic gap between low-level image features and high-level human knowledge, the image retrieval community has recently shifted its emphasis from low-level features analysis to high-level image semantics extrac- tion. User studies reveal that users tend to seek information using high-level semantics. Therefore, image semantics extraction is of great importance to content-based image retrieval because it allows the users to freely express what images they want. Semantic content annotation is the basis for semantic content retrieval. The aim of image anno- tation is to automatically obtain keywords that can be used to represent the content of images. The major research challenges in image semantic annotation are: what is the basic unit of semantic representation? how can the semantic unit be linked to high-level image knowledge? how can the contextual information be stored and utilized for image annotation? In this thesis, the Semantic Web technology (i.e. ontology) is introduced to the image semantic annotation problem. Semantic Web, the next generation web, aims at mak- ing the content of whatever type of media not only understandable to humans but also to machines. Due to the large amounts of multimedia data prevalent on the Web, re- searchers and industries are beginning to pay more attention to the Multimedia Semantic Web. The Semantic Web technology provides a new opportunity for multimedia-based applications, but the research in this area is still in its infancy. Whether ontology can be used to improve image annotation and how to best use ontology in semantic repre- sentation and extraction is still a worth-while investigation. This thesis deals with the problem of image semantic annotation using ontology and machine learning techniques in four phases as below. 1) Salient object extraction. A salient object servers as the basic unit in image semantic extraction as it captures the common visual property of the objects. Image segmen- tation is often used as the �rst step for detecting salient objects, but most segmenta- tion algorithms often fail to generate meaningful regions due to over-segmentation and under-segmentation. We develop a new salient object detection algorithm by combining multiple homogeneity criteria in a region merging framework. 2) Ontology construction. Since real-world objects tend to exist in a context within their environment, contextual information has been increasingly used for improving object recognition. In the ontology construction phase, visual-contextual ontologies are built from a large set of fully segmented and annotated images. The ontologies are composed of several types of concepts (i.e. mid-level and high-level concepts), and domain contextual knowledge. The visual-contextual ontologies stand as a user-friendly interface between low-level features and high-level concepts. 3) Image objects annotation. In this phase, each object is labelled with a mid-level concept in ontologies. First, a set of candidate labels are obtained by training Support Vectors Machines with features extracted from salient objects. After that, contextual knowledge contained in ontologies is used to obtain the �nal labels by removing the ambiguity concepts. 4) Scene semantic annotation. The scene semantic extraction phase is to get the scene type by using both mid-level concepts and domain contextual knowledge in ontologies. Domain contextual knowledge is used to create scene con�guration that describes which objects co-exist with which scene type more frequently. The scene con�guration is represented in a probabilistic graph model, and probabilistic inference is employed to calculate the scene type given an annotated image. To evaluate the proposed methods, a series of experiments have been conducted in a large set of fully annotated outdoor scene images. These include a subset of the Corel database, a subset of the LabelMe dataset, the evaluation dataset of localized semantics in images, the spatial context evaluation dataset, and the segmented and annotated IAPR TC-12 benchmark.

Monitoring mesenchymal stem cell cultures using image processing and pattern recognition techniques

Stem cells have attracted tremendous interest in recent times due to their promise in providing innovative new treatments for a great range of currently debilitating diseases. This is due to their potential ability to regenerate and repair damaged tissue, and hence restore lost body function, in a manner beyond the body's usual healing process. Bone marrow-derived mesenchymal stem cells or bone marrow stromal cells are one type of adult stem cells that are of particular interest. Since they are derived from a living human adult donor, they do not have the ethical issues associated with the use of human embryonic stem cells. They are also able to be taken from a patient or other donors with relative ease and then grown readily in the laboratory for clinical application. Despite the attractive properties of bone marrow stromal cells, there is presently no quick and easy way to determine the quality of a sample of such cells. Presently, a sample must be grown for weeks and subject to various time-consuming assays, under the direction of an expert cell biologist, to determine whether it will be useful. Hence there is a great need for innovative new ways to assess the quality of cell cultures for research and potential clinical application. The research presented in this thesis investigates the use of computerised image processing and pattern recognition techniques to provide a quicker and simpler method for the quality assessment of bone marrow stromal cell cultures. In particular, aim of this work is to find out whether it is possible, through the use of image processing and pattern recognition techniques, to predict the growth potential of a culture of human bone marrow stromal cells at early stages, before it is readily apparent to a human observer. With the above aim in mind, a computerised system was developed to classify the quality of bone marrow stromal cell cultures based on phase contrast microscopy images. Our system was trained and tested on mixed images of both healthy and unhealthy bone marrow stromal cell samples taken from three different patients. This system, when presented with 44 previously unseen bone marrow stromal cell culture images, outperformed human experts in the ability to correctly classify healthy and unhealthy cultures. The system correctly classified the health status of an image 88% of the time compared to an average of 72% of the time for human experts. Extensive training and testing of the system on a set of 139 normal sized images and 567 smaller image tiles showed an average performance of 86% and 85% correct classifications, respectively. The contributions of this thesis include demonstrating the applicability and potential of computerised image processing and pattern recognition techniques to the task of quality assessment of bone marrow stromal cell cultures. As part of this system, an image normalisation method has been suggested and a new segmentation algorithm has been developed for locating cell regions of irregularly shaped cells in phase contrast images. Importantly, we have validated the efficacy of both the normalisation and segmentation method, by demonstrating that both methods quantitatively improve the classification performance of subsequent pattern recognition algorithms, in discriminating between cell cultures of differing health status. We have shown that the quality of a cell culture of bone marrow stromal cells may be assessed without the need to either segment individual cells or to use time-lapse imaging. Finally, we have proposed a set of features, that when extracted from the cell regions of segmented input images, can be used to train current state of the art pattern recognition systems to predict the quality of bone marrow stromal cell cultures earlier and more consistently than human experts.

Machine recognition of hand-drawn circuit diagrams

An application of image processing techniques to recognition of hand-drawn circuit diagrams is presented. The scanned image of a diagram is pre-processed to remove noise and converted to bilevel. Morphological operations are applied to obtain a clean, connected representation using thinned lines. The diagram comprises of nodes, connections and components. Nodes and components are segmented using appropriate thresholds on a spatially varying object pixel density. Connection paths are traced using a pixel-stack. Nodes are classified using syntactic analysis. Components are classified using a combination of invariant moments, scalar pixel-distribution features, and vector relationships between straight lines in polygonal representations. A node recognition accuracy of 82% and a component recognition accuracy of 86% was achieved on a database comprising 107 nodes and 449 components. This recogniser can be used for layout “beautification” or to generate input code for circuit analysis and simulation packages

Particulate texture image analysis with applications

Texture analysis and textural cues have been applied for image classification, segmentation and pattern recognition. Dominant texture descriptors include directionality, coarseness, line-likeness etc. In this dissertation a class of textures known as particulate textures are defined, which are predominantly coarse or blob-like. The set of features that characterise particulate textures are different from those that characterise classical textures. These features are micro-texture, macro-texture, size, shape and compaction. Classical texture analysis techniques do not adequately capture particulate texture features. This gap is identified and new methods for analysing particulate textures are proposed. The levels of complexity in particulate textures are also presented ranging from the simplest images where blob-like particles are easily isolated from their back- ground to the more complex images where the particles and the background are not easily separable or the particles are occluded. Simple particulate images can be analysed for particle shapes and sizes. Complex particulate texture images, on the other hand, often permit only the estimation of particle dimensions. Real life applications of particulate textures are reviewed, including applications to sedimentology, granulometry and road surface texture analysis. A new framework for computation of particulate shape is proposed. A granulometric approach for particle size estimation based on edge detection is developed which can be adapted to the gray level of the images by varying its parameters. This study binds visual texture analysis and road surface macrotexture in a theoretical framework, thus making it possible to apply monocular imaging techniques to road surface texture analysis. Results from the application of the developed algorithm to road surface macro-texture, are compared with results based on Fourier spectra, the auto- correlation function and wavelet decomposition, indicating the superior performance of the proposed technique. The influence of image acquisition conditions such as illumination and camera angle on the results was systematically analysed. Experimental data was collected from over 5km of road in Brisbane and the estimated coarseness along the road was compared with laser profilometer measurements. Coefficient of determination R2 exceeding 0.9 was obtained when correlating the proposed imaging technique with the state of the art Sensor Measured Texture Depth (SMTD) obtained using laser profilometers.

A constraint to improve the reliability of stereo matching using the rank transform

The rank transform is a non-parametric technique which has been recently proposed for the stereo matching problem. The motivation behind its application to the matching problem is its invariance to certain types of image distortion and noise, as well as its amenability to real-time implementation. This paper derives an analytic expression for the process of matching using the rank transform, and then goes on to derive one constraint which must be satisfied for a correct match. This has been dubbed the rank order constraint or simply the rank constraint. Experimental work has shown that this constraint is capable of resolving ambiguous matches, thereby improving matching reliability. This constraint was incorporated into a new algorithm for matching using the rank transform. This modified algorithm resulted in an increased proportion of correct matches, for all test imagery used.

Porosity of serially sectioned interplanetary dust particles

The application of epoxy embedding and microtomy to individual chondritic interplanetary dust particles (lOP's)(Bradley and Brownlee, 1986a) provides not only higher precision in thin-film elemental analyses (Bradley and Brownlee, 19861:1), but also allows a wealth of other important techniques for the micro-characterization of these primitive extraterrestrial materials. For example, individual sections (e.g. 100 nm thick) or a series of sections, can be examined using image analysis techniques which utilize either transmitted or scanned secondary electron images, or alternatively, secondary X-ray spectra collected concurrently from a given region of sample. Individual particles, or groups of particles with similar image characteristics can then be rapidly identified using conventional grey-scale/particle recognition techniques for each microtomed section of lOP. This type of image analysis provides a suitable method for determination of particle size and shape distribution as well as porosity throughout the aggregate.

Image Region Salience for Improving Appearance-Based Place Recognition using a Supervised Classifier System

Many state of the art vision-based Simultaneous Localisation And Mapping (SLAM) and place recognition systems compute the salience of visual features in their environment. As computing salience can be problematic in radically changing environments new low resolution feature-less systems have been introduced, such as SeqSLAM, all of which consider the whole image. In this paper, we implement a supervised classifier system (UCS) to learn the salience of image regions for place recognition by feature-less systems. SeqSLAM only slightly benefits from the results of training, on the challenging real world Eynsham dataset, as it already appears to filter less useful regions of a panoramic image. However, when recognition is limited to specific image regions performance improves by more than an order of magnitude by utilising the learnt image region saliency. We then investigate whether the region salience generated from the Eynsham dataset generalizes to another car-based dataset using a perspective camera. The results suggest the general applicability of an image region salience mask for optimizing route-based navigation applications.