No description available.

No description available.

No description available.

Image and video enhancement, a classical problem of signal processing, has remained a very active research topic for past decades. This technical subject will not become obsolete even as the sensitivity and quality of modern image sensors steadily improve. No matter what level of sophistication cameras reach, there will always be more extreme and complex lighting conditions, in which the acquired images are improperly exposed and thus need to be enhanced. The central theme of enhancement is to algorithmically compensate for sensor limitations under ill lighting and make illegible details conspicuous, while maintaining a degree of naturalness. In retrospect, all existing contrast enhancement methods focus on heightening of spatial details in the luminance channel to fulfil the goal, with no or little consideration of the colour fidelity of the processed images; as a result they can introduce highly noticeable distortions in chrominance. This long-time much overlooked problem is addressed and systematically investigated by the thesis. We then propose a novel optimization-based enhancement algorithm, generating optimal tone mapping that not only makes maximal gain of contrast but also constrains tone and chrominance distortion, achieving superior output perceptual quality against severe underexposure and/or overexposure. Besides, we present a novel solution to restore images captured under more challenging backlit scenes, by combining the above enhancement method and feature-driven, machine learning based segmentation. We demonstrate the superior performance of the proposed method in terms of segmentation accuracy and restoration results over state-of-the-art methods. We also shed light on a common yet largely untreated video restoration problem called Yin-Yang Phasing (YYP), featured by involuntary, intense fluctuation in intensity and chrominance of an object as the video plays. We propose a novel video restoration technique to suppress YYP artifacts while retaining temporal consistency of objects appearance via inter-frame, spatially-adaptive optimal tone mapping. Experimental results are encouraging, pointing to an effective and practical solution to the problem. / Thesis / Doctor of Philosophy (PhD)

This paper introduces 3D-DDnet, a cutting-edge 3D deep learning (DL) framework designed to improve the image quality of low-dose computed tomography (LDCT) scans. Although LDCT scans are advantageous for reducing radiation exposure, they inherently suffer from reduced image quality. Our novel 3D DL architecture addresses this issue by effectively enhancing LDCT images to achieve parity with the quality of standard-dose CT scans. By exploiting the inter-slice correlation present in volumetric CT data, 3D-DDnet surpasses existing denoising benchmarks. It incorporates distributed data parallel (DDP) and transfer learning techniques to significantly accelerate the training process. The DDP approach is particularly tailored for operation across multiple Nvidia A100 GPUs, facilitating the processing of large-scale volumetric data sets that were previously unmanageable due to size constraints. Comparative analyses demonstrate that 3D-DDnet reduces the mean square error (MSE) by 10% over its 2D counterpart, 2D-DDnet. Moreover, by applying transfer learning from pre-trained 2D models, 3D-DDnet effectively 'jump starts' the learning process, cutting training times by half without compromising on model accuracy. / Master of Science / This research focuses on improving the quality of low-dose CT scans using advanced technology. CT scans are medical imaging techniques used to see inside the body. Low-dose CT (LDCT) scans use less radiation than standard CT scans, making them safer, but the downside is that the images are not as clear. To solve this problem, we developed a new deep learning method to make these low-dose images clearer and as good as regular CT scans. Our approach, called 3D-DDnet, is unique because it looks at the scans in 3D, considering how slices of the scan are related, which helps remove the noise and improve the image quality. Additionally, we used a technique called distributed data parallel (DDP) with advanced GPUs (graphics processing units, which are powerful computer components) to speed up the training of our system. This means our method can learn to improve images faster and work with larger data sets than before. Our results are promising: 3D-DDnet improved the image quality of low-dose CT scans significantly better than previous methods. Also, by using what we call "transfer learning" (starting with a pre-made model and adapting it), we cut the training time in half without losing accuracy. This development is essential for making low-dose CT scans more effective and safer for patients.

Yes / Digital image processing used in modern cardiac interventional x-ray systems may have the potential to enhance image quality such that it allows for lower radiation doses. The aim of this research was to quantify the reduction in radiation dose facilitated by image processing alone for percutaneous coronary intervention (PCI) patient angiograms, without reducing the perceived image quality required to confidently make a diagnosis. Incremental amounts of image noise were added to five PCI patient angiograms, simulating the angiogram having been acquired at corresponding lower dose levels (by 10-89% dose reduction). Sixteen observers with relevant and experience scored the image quality of these angiograms in three states - with no image processing and with two different modern image processing algorithms applied; these algorithms are used on state-of-the-art and previous generation cardiac interventional x-ray systems. Ordinal regression allowing for random effects and the delta method were used to quantify the dose reduction allowed for by the processing algorithms, for equivalent image quality scores. The dose reductions [with 95% confidence interval] from the state-of-the-art and previous generation image processing relative to no processing were 24.9% [18.8- 31.0%] and 15.6% [9.4-21.9%] respectively. The dose reduction enabled by the state-of-the-art image processing relative to previous generation processing was 10.3% [4.4-16.2%]. This demonstrates that statistically significant dose reduction can be facilitated with no loss in perceived image quality using modern image enhancement; the most recent processing algorithm was more effective in preserving image quality at lower doses. / The study was funded by Philips Healthcare (the Netherlands).

Prior to 1960's, the fingerprint analysis was carried out manually by human experts and for forensic purposes only. Automated fingerprint identification systems (AFIS) have been developed during the last 50 years. The success of AFIS resulted in that its use expanded beyond forensic applications and became common also in civilian applications. Mobile phones and computers equipped with fingerprint sensing devices for fingerprint-based user identification are common today. Despite the intense development efforts, a major problem in automatic fingerprint identification is to acquire reliable matching features from fingerprint images with poor quality. Images where the fingerprint pattern is heavily degraded usually inhibit the performance of an AFIS system. The performance of AFIS systems is also reduced when matching fingerprints of individuals with large age variations. This doctoral thesis presents contributions within the field of fingerprint image enhancement, segmentation and minutiae detection. The reliability of the extracted fingerprint features is highly dependent on the quality of the obtained fingerprints. Unfortunately, it is not always possible to have access to high quality fingerprints. Therefore, prior to the feature extraction, an enhancement of the quality of fingerprints and a segmentation are performed. The segmentation separates the fingerprint pattern from the background and thus limits possible sources of error due to, for instance, feature outliers. Most enhancement and segmentation techniques are data-driven and therefore based on certain features extracted from the low quality fingerprints at hand. Hence, different types of processing, such as directional filtering, are employed for the enhancement. This thesis contributes by proposing new research both for improving fingerprint matching and for the required pre-processing that improves the extraction of features to be used in fingerprint matching systems. In particular, the majority of enhancement and segmentation methods proposed herein are adaptive to the characteristics of each fingerprint image. Thus, the methods are insensitive towards sensor and fingerprint variability. Furthermore, introduction of the higher order statistics (kurtosis) for fingerprint segmentation is presented. Segmentation of the fingerprint image reduces the computational load by excluding background regions of the fingerprint image from being further processed. Also using a neural network to obtain a more robust minutiae detector with a patch rejection mechanism for speeding up the minutiae detection is presented in this thesis.

Le but de cette étude est d’évaluer la faisabilité de l’analyse tomodensitométrique 3D des bioprothèses aortiques pour faciliter leur évaluation morphologique durant le suivi et d’aider la sélection de cas et améliorer la planification d’une procédure valvein-valve. Le challenge était représenté par le rehaussement des feuillets valvulaires, en raison d’images très bruitées. Un angio-scanner synchronisé était réalisé chez des patients porteurs d’une bioprotèses aortique dégénérée avant réintervention (images in-vivo). Différentes méthodes pour la réduction du bruit étaient proposées. La reconstruction tridimensionnelle des bioprothèses était réalisée en utilisant des méthodes de segmentation de régions par "sticks". Après réopération ces méthodes étaient appliquées aux images scanner des bioprothèses explantées (images ex-vivo) et utilisées comme référence. La réduction du bruit obtenue par le filtre stick modifié montrait meilleurs résultats en rapport signal/bruit en comparaison aux filtres de diffusion anisotropique. Toutes les méthodes de segmentation ont permis une reconstruction 3D des feuillets. L’analyse qualitative a montré une bonne concordance entre les images obtenues in-vivo et les altérations des bioprothèses. Les résultats des différentes méthodes étaient comparés par critères volumétriques et discutés. Les bases d'une première approche de visualisation spatio-temporelle d'images TDM 3D+T de la prothèse valvulaire ont été proposés. Elle implique des techniques de rendu volumique et de compensation de mouvement. Son application à la valve native a aussi été envisagée. Les images scanner des bioprothèses aortiques nécessitent un traitement de débruitage et de réduction des artéfacts de façon à permettre le rehaussement des feuillets prothétiques. Les méthodes basées sticks semblent constituer une approche pertinente pour caractériser morphologiquement la dégénérescence des bioprothèses. / The aim of the study was to assess the feasibility of CT based 3D analysis of degenerated aortic bioprostheses to make easier their morphological assessment. This could be helpful during regular follow-up and for case selection, improved planning and mapping of valve-in-valve procedure. The challenge was represented by leaflets enhancement because of highly noised CT images. Contrast-enhanced ECG-gated CT scan was performed in patients with degenerated aortic bioprostheses before reoperation (in-vivo images). Different methods for noise reduction were tested and proposed. 3D reconstruction of bioprostheses components was achieved using stick based region segmentation methods. After reoperation, segmentation methods were applied to CT images of the explanted prostheses (exvivo images). Noise reduction obtained by improved stick filter showed best results in terms of signal to noise ratio comparing to anisotropic diffusion filters. All segmentation methods applied to the best phase of in-vivo images allowed 3D bioprosthetic leaflets reconstruction. Explanted bioprostheses CT images were also processed and used as reference. Qualitative analysis revealed a good concordance between the in-vivo images and the bioprostheses alterations. Results from different methods were compared by means of volumetric criteria and discussed. A first approach for spatiotemporal visualization of 3D+T images of valve bioprosthesis has been proposed. Volume rendering and motion compensation techniques were applied to visualize different phases of CT data. Native valve was also considered. ECG-gated CT images of aortic bioprostheses need a preprocessing to reduce noise and artifacts in order to enhance prosthetic leaflets. Stick based methods seems to provide an interesting approach for the morphological characterization of degenerated bioprostheses.

In this thesis, opti-acoustic stereo imaging, which is the deployment of two-dimensional (2D) high frequency imaging sonar with the electro-optical camera in calibrated stereo configuration, is studied. Optical cameras give detailed images in clear waters. However, in dark or turbid waters, information coming from electro-optical sensor is insufficient for accurate scene perception. Imaging sonars, also known as acoustic cameras, can provide enhanced target details under these scenarios. To illustrate these visibility conditions, a 2D high frequency imaging sonar simulator as well as an underwater optical image simulator is developed. A computationally efficient algorithm is also proposed for the post-processing of the returned sonar signals. Where optical visibility allows, integration of the sonar and optical images effectively provides binocular stereo vision capability and enables the recovery of three-dimensional (3D) structural information. This requires solving the feature correspondence problem for these completely different sensing modalities. Geometrical interpretation of this problem is examined on the simulated optical and sonar images. Matching the features manually, 3D reconstruction performance of opti-acoustic system is also investigated. In addition, motion estimation from opti-acoustic image sequences is studied. Finally, a method is proposed to improve the degraded optical images with the help of sonar images. First, a nonlinear mapping is found to match local the features in opti-acoustical images. Next, features in the sonar image is mapped to the optical image using the transformation. Performance of the mapping is evaluated for different scene geometries.

The presence of fractures, joints and karst features within rock strongly influence the hydraulic and mechanical behavior of a rock mass, and there is a strong desire to characterize these features in a noninvasive manner, such as by using ground penetrating radar (GPR). These features can alter the incident waveform and polarization of the GPR signal depending on the aperture, fill and orientation of the features. The GPR methods developed here focus on changes in waveform, polarization or texture that can improve the detection and discrimination of these features within rock bodies. These new methods are utilized to better understand the interaction of an invasive shrub, Juniperus ashei, with subsurface flow conduits at an ecohydrologic experimentation plot situated on the limestone of the Edwards Aquifer, central Texas. First, a coherency algorithm is developed for polarimetric GPR that uses the largest eigenvalue of a scattering matrix in the calculation of coherence. This coherency is sensitive to waveshape and unbiased by the polarization of the GPR antennas, and it shows improvement over scalar coherency in detection of possible conduits in the plot data. Second, a method is described for full-waveform inversion of transmission data to quantitatively determine fracture aperture and electromagnetic properties of the fill, based on a thin-layer model. This inversion method is validated on synthetic data, and the results from field data at the experimentation plot show consistency with the reflection data. Finally, growing hierarchical self-organizing maps (GHSOM) are applied to the GPR data to discover new patterns indicative of subsurface features, without representative examples. The GHSOMs are able to distinguish patterns indicating soil filled cavities within the limestone. Using these methods, locations of soil filled cavities and the dominant flow conduits were indentified. This information helps to reconcile previous hydrologic experiments conducted at the site. Additionally, the GPR and hydrologic experiments suggests that Juniperus ashei significantly impacts infiltration by redirecting flow towards its roots occupying conduits and soil bodies within the rock. This research demonstrates that GPR provides a noninvasive tool that can improve future subsurface experimentation.