Exemplar-based learning or, equally, nearest neighbor methods have recently gained interest from researchers in a variety of computer science domains because of the prevalence of large amounts of accessible data and storage capacity. In computer vision, these types of technique have been successful in several problems such as scene recognition, shape matching, image parsing, character recognition, and object detection. Applying the concept of exemplar-based learning to the problem of color constancy seems odd at first glance since, in the first place, similar nearest neighbor images are not usually affected by precisely similar illuminants and, in the second place, gathering a dataset consisting of all possible real-world images, including indoor and outdoor scenes and for all possible illuminant colors and intensities, is indeed impossible. In this paper, we instead focus on surfaces in the image and address the color constancy problem by unsupervised learning of an appropriate model for each training surface in training images. We find nearest neighbor models for each surface in a test image and estimate its illumination based on comparing the statistics of pixels belonging to nearest neighbor surfaces and the target surface. The final illumination estimation results from combining these estimated illuminants over surfaces to generate a unique estimate. We show that it performs very well, for standard datasets, compared to current color constancy algorithms, including when learning based on one image dataset is applied to tests from a different dataset. The proposed method has the advantage of overcoming multi-illuminant situations, which is not possible for most current methods since they assume the color of the illuminant is constant all over the image. We show a technique to overcome the multiple illuminant situation using the proposed method and test our technique on images with two distinct sources of illumination using a multiple-illuminant color constancy dataset. The concept proposed here is a completely new approach to the color constancy problem and provides a simple learning-based framework.

Exemplar-Based Color Constancy and Multiple Illumination

approved: ______________________________________________________ T. Matthew Evans Anchoring systems are used to hold floating offshore devices (e.g., energy devices, oil platforms) to keep them on their stations. Anchors are embedded into the seabed soils. In some cases, the interface shear between anchor and seabed soils together with anchor self-weight provide the holding capacity of the anchoring systems. However, in other cases, the embedment behavior (e.g., embedment depth) and the geometry of anchor play significant roles on the holding capacity. Anchors embedded into soils are soil-structure interaction problems, which have been widely studied experimentally, analytically, and numerically. Results from experimental tests, analytical and numerical analyses have been reported in the literature looking for the macroscale response. However, few studies focus on looking for the microscale insight of the anchor-soil responses. The discrete element method (DEM) and digital image processing (DIP) can be used for microscale analysis which has been barely applied on the soil-structure practices. In the dissertation, DEM and DIP are used to investigate the macroand microscale soil-anchor interactions in four chapters. Chapter 3 describes DIP and its application to characterize granular assemblies quantitatively and morphologically. The investigation ranges from obtaining the center of mass of particles, sphericity, local void ratio distribution, moment of inertia, and particle orientation. The void space and particle characteristics have a significant influence on the macroscale response of a granular assembly. Accurately quantification of void space and identification of particle characteristics will give better understandings of granular responses. This chapter presents quantitative approaches of local void space quantification and particle characteristics identification in the aid of three-dimensional digital image analysis. As an in-depth analysis, the authors deepen Oda’s two-dimensional image segmentation approach to three-dimensional by using Delaunay triangulation. Morphological processing (e.g. image erosion, dilation and opening) of threedimensional binary images is used for both image segmentation and particle characteristic identification. Algorithms of void space quantification, particle characteristics identification and particle seeking are all built as well. The segmentation approach has been verified by applying it to regular packings: simple cubic and tetrahedral. Particle characteristic identification and particle seeking algorithms were verified by application to particles/assemblies with known information (e.g. geometry). Results show that the algorithms are robust and can accurately quantify/identify void space/particle characteristics of the granular assembly. Chapter 4 presents a granular-continuum interface shear model using DEM and investigates the microscale behaviors of the interface. Granular-continuum interfaces are widely present in geotechnical structures, including deep foundations, retaining structures, and anchoring applications. Interface mechanical properties are a function of the properties of the contacting soil and the characteristics of the opposing interface. Therefore, a robust understanding of granularcontinuum interface behavior is essential to geotechnical practice. This chapter provides a summary of recent research on the effects of interface roughness, soil density, particle shape, and friction coefficient on interface behavior and strength and results from three-dimensional DEM simulations of granular-continuum interface shear. The trends in DEM results are compared to the previously published physical experiments; and the microscale responses of the interface simulations has been investigated. The results show that there are good agreements between DEM simulation and experimental tests under similar interface roughnesses. DEM simulations give a similar bilinear strength-displacement trend as that previously reported from physical experiments. The interface failure mechanism is contact reorientation for rougher interfaces and contact sliding for smoother interfaces, as shown through microscale investigations. Mobilization of rougher interfaces tend to change to force distribution of the surrounding granular soils. In Chapter 5, the response of plate anchors under quasi-static and cyclic loading conditions is considered. Plate anchors are embedded into the ocean floor to provide holding capacity for offshore structures. Anchor holding capacity is a function of both the anchor and soil properties. Although plate anchors have been widely studied experimentally and numerically, there is still no universally agreed-upon design approach, indicating that the problem physics remain elusive. In this work, DEM simulations are used to investigate the behavior of plate anchors during pullout in an effort to elucidate some of the microscale physical processes that influence overall system behavior. Macroscale assembly response is compared to published experimental results and empirical solutions. The influence of embedment ratio, anchor roughness, soil density, and anchor size on holding capacity are investigated and system-scale results are shown to reasonably agree with previously published work. Thus, observations of the simulated contact force network and particle velocity during uplift are used to provide insight into anchor failure mechanisms. Finally, the model is used to briefly explore the response of a cyclically-loaded plate anchor embedded in a granular assembly. Results from DEM simulations of torpedo anchor penetration and the associated soil response are presented in Chapter 6. Torpedo anchors are a viable approach for mooring marine hydrokinetic (MHK) energy devices to the seafloor. These anchors can serve to maintain station and to provide the reaction force for an MHK device. The ability of the anchor to perform these duties is a strong function of its penetration depth during installation. This is a large-strain problem not amenable to typical continuum numerical approaches. In the current work, we propose that the discrete element method (DEM) is a more appropriate tool to investigate the shallow penetration of torpedo anchors in sands. The effects of anchor mass, impact velocity, and soil interparticle friction are considered in the DEM simulations. The relative maximum penetration depths for different penetration conditions are quantified and presented. Granular material response at the microscale during penetration are used to provide insight into system response. Energy dissipation in the assembly by both friction and collision at the particle scale are considered. Results show that anchor penetration increases approximately linearly with an increase in impact velocity or anchor weight. Penetration decreases with an increase in interparticle friction (i.e., soil strength). Observations of microscale behaviors and energy calculations are used to provide insight into overall system response. ©Copyright by Nan Zhang April 3, 2018 All Rights Reserved NUMERICAL SIMULATIONS AND MICROSCALE ANALYSES OF OFFSHORE ANCHORGRANULAR MATERIAL SYSTEMS

Abstract of dissertation

The Zeta-image, illuminant estimation, and specularity manipulation

Splicing operation is a common technique used for image forgery. In this paper, we propose a new method for detecting the image splicing by revealing inconsistencies of the illuminant color in the object regions. For this, we first divide a given image into some horizontal and vertical bands. After that, the illuminant of each band is estimated with the generalized grey-world algorithms. For each illuminant estimation algorithm, the fake patches produced by computing the intersection between the forged horizontal and vertical bands are represented in a forgery detection map. Finally the spliced regions are shown by combining all splicing detection maps. Our proposed method is validated with several test cases. Additionally, we compare the performance of our method with two existing methods on the dataset CASIA V2.0. The results show that the proposed method works well with minimal human interaction.

Image splicing detection with local illumination estimation

High dynamic range (HDR) imaging seems to have developed to a level of soon being a standard feature in
consumer cameras. This study was motivated by the need for evaluating tone mapping operators especially
for consumer imaging applications. A no-reference method based on ISO 20462-2:2005 triplet comparison was
created for evaluating tone mapping operators. Multiple HDR test images were photographed and the method
was validated by evaluating 25 tone mapping operators with five test images. Tone mapping operators were
evaluated based on image naturalness and pleasantness. The results indicate that the method successfully ranked
the method in terms of naturalness and pleasantness. The test image set could be improved for example based
on an imaging photo space for HDR photography. The test images of this study are available for non-commercial
research purposes.

Weihua Xiong

Papers

Color Constancy for Multi-Illuminants High-Dynamic-Range Scenes.