Regression analysis of ranked segment parameters for optic nerve head classification:a pilot study

Purpose: To determine whether curve-fitting analysis of the ranked segment distributions of topographic optic nerve head (ONH) parameters, derived using the Heidelberg Retina Tomograph (HRT), provide a more effective statistical descriptor to differentiate the normal from the glaucomatous ONH. Methods: The sample comprised of 22 normal control subjects (mean age 66.9 years; S.D. 7.8) and 22 glaucoma patients (mean age 72.1 years; S.D. 6.9) confirmed by reproducible visual field defects on the Humphrey Field Analyser. Three 10°-images of the ONH were obtained using the HRT. The mean topography image was determined and the HRT software was used to calculate the rim volume, rim area to disc area ratio, normalised rim area to disc area ratio and retinal nerve fibre cross-sectional area for each patient at 10°-sectoral intervals. The values were ranked in descending order, and each ranked-segment curve of ordered values was fitted using the least squares method. Results: There was no difference in disc area between the groups. The group mean cup-disc area ratio was significantly lower in the normal group (0.204 ± 0.16) compared with the glaucoma group (0.533 ± 0.083) (p < 0.001). The visual field indices, mean deviation and corrected pattern S.D., were significantly greater (p < 0.001) in the glaucoma group (-9.09 dB ± 3.3 and 7.91 ± 3.4, respectively) compared with the normal group (-0.15 dB ± 0.9 and 0.95 dB ± 0.8, respectively). Univariate linear regression provided the best overall fit to the ranked segment data. The equation parameters of the regression line manually applied to the normalised rim area-disc area and the rim area-disc area ratio data, correctly classified 100% of normal subjects and glaucoma patients. In this study sample, the regression analysis of ranked segment parameters method was more effective than conventional ranked segment analysis, in which glaucoma patients were misclassified in approximately 50% of cases. Further investigation in larger samples will enable the calculation of confidence intervals for normality. These reference standards will then need to be investigated for an independent sample to fully validate the technique. Conclusions: Using a curve-fitting approach to fit ranked segment curves retains information relating to the topographic nature of neural loss. Such methodology appears to overcome some of the deficiencies of conventional ranked segment analysis, and subject to validation in larger scale studies, may potentially be of clinical utility for detecting and monitoring glaucomatous damage. © 2007 The College of Optometrists.

Adaptive Document Image Binarization with Application in Processing Astronomical Logbooks

ACM Computing Classification System (1998): I.7, I.7.5.

Modular Digital Watermarking for Image Verification and Secure Data Storage in Web Applications

Our modular approach to data hiding is an innovative concept in the data hiding research field. It enables the creation of modular digital watermarking methods that have extendable features and are designed for use in web applications. The methods consist of two types of modules – a basic module and an application-specific module. The basic module mainly provides features which are connected with the specific image format. As JPEG is a preferred image format on the Internet, we have put a focus on the achievement of a robust and error-free embedding and retrieval of the embedded data in JPEG images. The application-specific modules are adaptable to user requirements in the concrete web application. The experimental results of the modular data watermarking are very promising. They indicate excellent image quality, satisfactory size of the embedded data and perfect robustness against JPEG transformations with prespecified compression ratios. ACM Computing Classification System (1998): C.2.0.

Circumarctic wetland dataset based on ENVISAT ASAR Global Monitoring Mode with links to geotiff image and NetCDF files (2005-01 to 2011-12)

A circumpolar representative and consistent wetland map is required for a range of applications ranging from upscaling of carbon fluxes and pools to climate modelling and wildlife habitat assessment. Currently available data sets lack sufficient accuracy and/or thematic detail in many regions of the Arctic. Synthetic aperture radar (SAR) data from satellites have already been shown to be suitable for wetland mapping. Envisat Advanced SAR (ASAR) provides global medium-resolution data which are examined with particular focus on spatial wetness patterns in this study. It was found that winter minimum backscatter values as well as their differences to summer minimum values reflect vegetation physiognomy units of certain wetness regimes. Low winter backscatter values are mostly found in areas vegetated by plant communities typically for wet regions in the tundra biome, due to low roughness and low volume scattering caused by the predominant vegetation. Summer to winter difference backscatter values, which in contrast to the winter values depend almost solely on soil moisture content, show expected higher values for wet regions. While the approach using difference values would seem more reasonable in order to delineate wetness patterns considering its direct link to soil moisture, it was found that a classification of winter minimum backscatter values is more applicable in tundra regions due to its better separability into wetness classes. Previous approaches for wetland detection have investigated the impact of liquid water in the soil on backscatter conditions. In this study the absence of liquid water is utilized. Owing to a lack of comparable regional to circumpolar data with respect to thematic detail, a potential wetland map cannot directly be validated; however, one might claim the validity of such a product by comparison with vegetation maps, which hold some information on the wetness status of certain classes. It was shown that the Envisat ASAR-derived classes are related to wetland classes of conventional vegetation maps, indicating its applicability; 30% of the land area north of the treeline was identified as wetland while conventional maps recorded 1-7%.

Statistical methods for polyhedral shape classification with incomplete data - application to cryo-electron tomographic images

A certain type of bacterial inclusion, known as a bacterial microcompartment, was recently identified and imaged through cryo-electron tomography. A reconstructed 3D object from single-axis limited angle tilt-series cryo-electron tomography contains missing regions and this problem is known as the missing wedge problem. Due to missing regions on the reconstructed images, analyzing their 3D structures is a challenging problem. The existing methods overcome this problem by aligning and averaging several similar shaped objects. These schemes work well if the objects are symmetric and several objects with almost similar shapes and sizes are available. Since the bacterial inclusions studied here are not symmetric, are deformed, and show a wide range of shapes and sizes, the existing approaches are not appropriate. This research develops new statistical methods for analyzing geometric properties, such as volume, symmetry, aspect ratio, polyhedral structures etc., of these bacterial inclusions in presence of missing data. These methods work with deformed and non-symmetric varied shaped objects and do not necessitate multiple objects for handling the missing wedge problem. The developed methods and contributions include: (a) an improved method for manual image segmentation, (b) a new approach to 'complete' the segmented and reconstructed incomplete 3D images, (c) a polyhedral structural distance model to predict the polyhedral shapes of these microstructures, (d) a new shape descriptor for polyhedral shapes, named as polyhedron profile statistic, and (e) the Bayes classifier, linear discriminant analysis and support vector machine based classifiers for supervised incomplete polyhedral shape classification. Finally, the predicted 3D shapes for these bacterial microstructures belong to the Johnson solids family, and these shapes along with their other geometric properties are important for better understanding of their chemical and biological characteristics.

Using Image Processing Methods to Improve the Detection of Buried Explosive Threats in GPR Data

Current state of the art techniques for landmine detection in ground penetrating radar (GPR) utilize statistical methods to identify characteristics of a landmine response. This research makes use of 2-D slices of data in which subsurface landmine responses have hyperbolic shapes. Various methods from the field of visual image processing are adapted to the 2-D GPR data, producing superior landmine detection results. This research goes on to develop a physics-based GPR augmentation method motivated by current advances in visual object detection. This GPR specific augmentation is used to mitigate issues caused by insufficient training sets. This work shows that augmentation improves detection performance under training conditions that are normally very difficult. Finally, this work introduces the use of convolutional neural networks as a method to learn feature extraction parameters. These learned convolutional features outperform hand-designed features in GPR detection tasks. This work presents a number of methods, both borrowed from and motivated by the substantial work in visual image processing. The methods developed and presented in this work show an improvement in overall detection performance and introduce a method to improve the robustness of statistical classification.

Automatically Evolving Rotation-invariant Texture Image Descriptors by Genetic Programming

In computer vision, training a model that performs classification effectively is highly dependent on the extracted features, and the number of training instances. Conventionally, feature detection and extraction are performed by a domain-expert who, in many cases, is expensive to employ and hard to find. Therefore, image descriptors have emerged to automate these tasks. However, designing an image descriptor still requires domain-expert intervention. Moreover, the majority of machine learning algorithms require a large number of training examples to perform well. However, labelled data is not always available or easy to acquire, and dealing with a large dataset can dramatically slow down the training process. In this paper, we propose a novel Genetic Programming based method that automatically synthesises a descriptor using only two training instances per class. The proposed method combines arithmetic operators to evolve a model that takes an image and generates a feature vector. The performance of the proposed method is assessed using six datasets for texture classification with different degrees of rotation, and is compared with seven domain-expert designed descriptors. The results show that the proposed method is robust to rotation, and has significantly outperformed, or achieved a comparable performance to, the baseline methods.

Mapping of Submerged Aquatic Vegetation in Rivers From Very High Resolution Image Data, Using Object Based Image Analysis Combined with Expert Knowledge

The use of remote sensing for monitoring of submerged aquatic vegetation (SAV) in fluvial environments has been limited by the spatial and spectral resolution of available image data. The absorption of light in water also complicates the use of common image analysis methods. This paper presents the results of a study that uses very high resolution (VHR) image data, collected with a Near Infrared sensitive DSLR camera, to map the distribution of SAV species for three sites along the Desselse Nete, a lowland river in Flanders, Belgium. Plant species, including Ranunculus aquatilis L., Callitriche obtusangula Le Gall, Potamogeton natans L., Sparganium emersum L. and Potamogeton crispus L., were classified from the data using Object-Based Image Analysis (OBIA) and expert knowledge. A classification rule set based on a combination of both spectral and structural image variation (e.g. texture and shape) was developed for images from two sites. A comparison of the classifications with manually delineated ground truth maps resulted for both sites in 61% overall accuracy. Application of the rule set to a third validation image, resulted in 53% overall accuracy. These consistent results show promise for species level mapping in such biodiverse environments, but also prompt a discussion on assessment of classification accuracy.

Classification multi-échelle d'images à très haute résolution spatiale basée sur une nouvelle approche texturale

Résumé : Face à l’accroissement de la résolution spatiale des capteurs optiques satellitaires, de nouvelles stratégies doivent être développées pour classifier les images de télédétection. En effet, l’abondance de détails dans ces images diminue fortement l’efficacité des classifications spectrales; de nombreuses méthodes de classification texturale, notamment les approches statistiques, ne sont plus adaptées. À l’inverse, les approches structurelles offrent une ouverture intéressante : ces approches orientées objet consistent à étudier la structure de l’image pour en interpréter le sens. Un algorithme de ce type est proposé dans la première partie de cette thèse. Reposant sur la détection et l’analyse de points-clés (KPC : KeyPoint-based Classification), il offre une solution efficace au problème de la classification d’images à très haute résolution spatiale. Les classifications effectuées sur les données montrent en particulier sa capacité à différencier des textures visuellement similaires. Par ailleurs, il a été montré dans la littérature que la fusion évidentielle, reposant sur la théorie de Dempster-Shafer, est tout à fait adaptée aux images de télédétection en raison de son aptitude à intégrer des concepts tels que l’ambiguïté et l’incertitude. Peu d’études ont en revanche été menées sur l’application de cette théorie à des données texturales complexes telles que celles issues de classifications structurelles. La seconde partie de cette thèse vise à combler ce manque, en s’intéressant à la fusion de classifications KPC multi-échelle par la théorie de Dempster-Shafer. Les tests menés montrent que cette approche multi-échelle permet d’améliorer la classification finale dans le cas où l’image initiale est de faible qualité. De plus, l’étude effectuée met en évidence le potentiel d’amélioration apporté par l’estimation de la fiabilité des classifications intermédiaires, et fournit des pistes pour mener ces estimations.

Caractérisation de minéraux indicateurs par imagerie hyperspectrale et traitement de l'image dans l'infrarouge proche et l'infrarouge lointain

L’imagerie hyperspectrale (HSI) fournit de l’information spatiale et spectrale concernant l’émissivité de la surface des matériaux, ce qui peut être utilisée pour l’identification des minéraux. Pour cela, un matériel de référence ou endmember, qui en minéralogie est la forme la plus pure d’un minéral, est nécessaire. L’objectif principal de ce projet est l’identification des minéraux par imagerie hyperspectrale. Les informations de l’imagerie hyperspectrale ont été enregistrées à partir de l’énergie réfléchie de la surface du minéral. L’énergie solaire est la source d’énergie dans l’imagerie hyperspectrale de télédétection, alors qu’un élément chauffant est la source d’énergie utilisée dans les expériences de laboratoire. Dans la première étape de ce travail, les signatures spectrales des minéraux purs sont obtenues avec la caméra hyperspectrale, qui mesure le rayonnement réfléchi par la surface des minéraux. Dans ce projet, deux séries d’expériences ont été menées dans différentes plages de longueurs d’onde (0,4 à 1 µm et 7,7 à 11,8 µm). Dans la deuxième partie de ce projet, les signatures spectrales obtenues des échantillons individuels sont comparées avec des signatures spectrales de la bibliothèque hyperspectrale de l’ASTER. Dans la troisième partie, trois méthodes différentes de classification hyperspectrale sont considérées pour la classification. Spectral Angle Mapper (SAM), Spectral Information Divergence (SID), et Intercorrélation normalisée (NCC). Enfin, un système d’apprentissage automatique, Extreme Learning Machine (ELM), est utilisé pour identifier les minéraux. Deux types d’échantillons ont été utilisés dans ce projet. Le système d’ELM est divisé en deux parties, la phase d’entraînement et la phase de test du système. Dans la phase d’entraînement, la signature d’un seul échantillon minéral est entrée dans le système, et dans la phase du test, les signatures spectrales des différents minéraux, qui sont entrées dans la phase d’entraînement, sont comparées par rapport à des échantillons de minéraux mixtes afin de les identifier.

Detection of mesoscale thermal fronts from 4km data using smoothing techniques: Gradient-based fronts classification and basin scale application

In order to optimize frontal detection in sea surface temperature fields at 4 km resolution, a combined statistical and expert-based approach is applied to test different spatial smoothing of the data prior to the detection process. Fronts are usually detected at 1 km resolution using the histogram-based, single image edge detection (SIED) algorithm developed by Cayula and Cornillon in 1992, with a standard preliminary smoothing using a median filter and a 3 × 3 pixel kernel. Here, detections are performed in three study regions (off Morocco, the Mozambique Channel, and north-western Australia) and across the Indian Ocean basin using the combination of multiple windows (CMW) method developed by Nieto, Demarcq and McClatchie in 2012 which improves on the original Cayula and Cornillon algorithm. Detections at 4 km and 1 km of resolution are compared. Fronts are divided in two intensity classes (“weak” and “strong”) according to their thermal gradient. A preliminary smoothing is applied prior to the detection using different convolutions: three type of filters (median, average and Gaussian) combined with four kernel sizes (3 × 3, 5 × 5, 7 × 7, and 9 × 9 pixels) and three detection window sizes (16 × 16, 24 × 24 and 32 × 32 pixels) to test the effect of these smoothing combinations on reducing the background noise of the data and therefore on improving the frontal detection. The performance of the combinations on 4 km data are evaluated using two criteria: detection efficiency and front length. We find that the optimal combination of preliminary smoothing parameters in enhancing detection efficiency and preserving front length includes a median filter, a 16 × 16 pixel window size, and a 5 × 5 pixel kernel for strong fronts and a 7 × 7 pixel kernel for weak fronts. Results show an improvement in detection performance (from largest to smallest window size) of 71% for strong fronts and 120% for weak fronts. Despite the small window used (16 × 16 pixels), the length of the fronts has been preserved relative to that found with 1 km data. This optimal preliminary smoothing and the CMW detection algorithm on 4 km sea surface temperature data are then used to describe the spatial distribution of the monthly frequencies of occurrence for both strong and weak fronts across the Indian Ocean basin. In general strong fronts are observed in coastal areas whereas weak fronts, with some seasonal exceptions, are mainly located in the open ocean. This study shows that adequate noise reduction done by a preliminary smoothing of the data considerably improves the frontal detection efficiency as well as the global quality of the results. Consequently, the use of 4 km data enables frontal detections similar to 1 km data (using a standard median 3 × 3 convolution) in terms of detectability, length and location. This method, using 4 km data is easily applicable to large regions or at the global scale with far less constraints of data manipulation and processing time relative to 1 km data.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.

Efficient Image and Video Representations for Retrieval

Image (Video) retrieval is an interesting problem of retrieving images (videos) similar to the query. Images (Videos) are represented in an input (feature) space and similar images (videos) are obtained by finding nearest neighbors in the input representation space. Numerous input representations both in real valued and binary space have been proposed for conducting faster retrieval. In this thesis, we present techniques that obtain improved input representations for retrieval in both supervised and unsupervised settings for images and videos. Supervised retrieval is a well known problem of retrieving same class images of the query. We address the practical aspects of achieving faster retrieval with binary codes as input representations for the supervised setting in the first part, where binary codes are used as addresses into hash tables. In practice, using binary codes as addresses does not guarantee fast retrieval, as similar images are not mapped to the same binary code (address). We address this problem by presenting an efficient supervised hashing (binary encoding) method that aims to explicitly map all the images of the same class ideally to a unique binary code. We refer to the binary codes of the images as `Semantic Binary Codes' and the unique code for all same class images as `Class Binary Code'. We also propose a new class­ based Hamming metric that dramatically reduces the retrieval times for larger databases, where only hamming distance is computed to the class binary codes. We also propose a Deep semantic binary code model, by replacing the output layer of a popular convolutional Neural Network (AlexNet) with the class binary codes and show that the hashing functions learned in this way outperforms the state­ of ­the art, and at the same time provide fast retrieval times. In the second part, we also address the problem of supervised retrieval by taking into account the relationship between classes. For a given query image, we want to retrieve images that preserve the relative order i.e. we want to retrieve all same class images first and then, the related classes images before different class images. We learn such relationship aware binary codes by minimizing the similarity between inner product of the binary codes and the similarity between the classes. We calculate the similarity between classes using output embedding vectors, which are vector representations of classes. Our method deviates from the other supervised binary encoding schemes as it is the first to use output embeddings for learning hashing functions. We also introduce new performance metrics that take into account the related class retrieval results and show significant gains over the state­ of­ the art. High Dimensional descriptors like Fisher Vectors or Vector of Locally Aggregated Descriptors have shown to improve the performance of many computer vision applications including retrieval. In the third part, we will discuss an unsupervised technique for compressing high dimensional vectors into high dimensional binary codes, to reduce storage complexity. In this approach, we deviate from adopting traditional hyperplane hashing functions and instead learn hyperspherical hashing functions. The proposed method overcomes the computational challenges of directly applying the spherical hashing algorithm that is intractable for compressing high dimensional vectors. A practical hierarchical model that utilizes divide and conquer techniques using the Random Select and Adjust (RSA) procedure to compress such high dimensional vectors is presented. We show that our proposed high dimensional binary codes outperform the binary codes obtained using traditional hyperplane methods for higher compression ratios. In the last part of the thesis, we propose a retrieval based solution to the Zero shot event classification problem - a setting where no training videos are available for the event. To do this, we learn a generic set of concept detectors and represent both videos and query events in the concept space. We then compute similarity between the query event and the video in the concept space and videos similar to the query event are classified as the videos belonging to the event. We show that we significantly boost the performance using concept features from other modalities.