Hyeokho Choi

ForWaRD: Fourier-wavelet regularized deconvolution for ill-conditioned systems

We propose an efficient, hybrid Fourier-wavelet regularized deconvolution (ForWaRD) algorithm that performs noise regularization via scalar shrinkage in both the Fourier and wavelet domains. The Fourier shrinkage exploits the Fourier transforms economical representation of the colored noise inherent in deconvolution, whereas the wavelet shrinkage exploits the wavelet domains economical representation of piecewise smooth signals and images. We derive the optimal balance between the amount of Fourier and wavelet regularization by optimizing an approximate mean-squared error (MSE) metric and find that signals with more economical wavelet representations require less Fourier shrinkage. ForWaRD is applicable to all ill-conditioned deconvolution problems, unlike the purely wavelet-based wavelet-vaguelette deconvolution (WVD); moreover, its estimate features minimal ringing, unlike the purely Fourier-based Wiener deconvolution. Even in problems for which the WVD was designed, we prove that ForWaRDs MSE decays with the optimal WVD rate as the number of samples increases. Further, we demonstrate that over a wide range of practical sample-lengths, ForWaRD improves on WVDs performance.

Multiscale image segmentation using wavelet-domain hidden Markov models

We introduce a new image texture segmentation algorithm, HMTseg, based on wavelets and the hidden Markov tree (HMT) model. The HMT is a tree-structured probabilistic graph that captures the statistical properties of the coefficients of the wavelet transform. Since the HMT is particularly well suited to images containing singularities (edges and ridges), it provides a good classifier for distinguishing between textures. Utilizing the inherent tree structure of the wavelet HMT and its fast training and likelihood computation algorithms, we perform texture classification at a range of different scales. We then fuse these multiscale classifications using a Bayesian probabilistic graph to obtain reliable final segmentations. Since HMTseg works on the wavelet transform of the image, it can directly segment wavelet-compressed images without the need for decompression into the space domain. We demonstrate the performance of HMTseg with synthetic, aerial photo, and document image segmentations.

Hidden Markov tree modeling of complex wavelet transforms

Multiresolution signal and image models such as the hidden Markov tree aim to capture the statistical structure of smooth and singular (edgy) regions. Unfortunately, models based on the orthogonal wavelet transform suffer from shift-variance, making them less accurate and realistic. We extend the HMT modeling framework to the complex wavelet transform, which features near shift-invariance and improved angular resolution compared to the standard wavelet transform. The model is computationally efficient (with linear-time computation and processing algorithms) and applicable to general Bayesian inference problems as a prior density for the data. In a simple estimation experiment, the complex wavelet HMT model outperforms a number of high-performance denoising algorithms, including redundant wavelet thresholding (cycle spinning) and the redundant HMT.

Multiscale document segmentation using wavelet-domain hidden Markov models

We introduce a new document image segmentation algorithm, HMTseg, based on wavelets and the hidden Markov tree (HMT) model. The HMT is a tree-structured probabilistic graph that captures the statistical properties of the coefficients of the wavelet transform. Since the HMT is particularly well suited to images containing singularities (edges and ridges), it provides a good classifier for distinguishing between different document textures. Utilizing the inherent tree structure of the wavelet HMT and its fast training and likelihood computation algorithms, we perform multiscale texture classification at a range of different scales. We then fuse these multiscale classifications using a Bayesian probabilistic graph to obtain reliable final segmentations. Since HMTseg works on the wavelet transform of the image, it can directly segment wavelet-compressed images, without the need for decompression into the space domain. We demonstrate HMTsegs performance with both synthetic and real imagery.

Analysis of wavelet-domain Wiener filters

We investigate Wiener filtering of wavelet coefficients for signal denoising. Empirically designed wavelet-domain Wiener filters outperform many other denoising algorithms based on wavelet thresholding. However, up to now, it has not been clear how to choose the signal model used to design the filter, because the effect of model selection on the filter performance is difficult to understand. By analyzing the error involved in the Wiener filter designed with an empirically obtained signal model, we show that hard thresholding is typically outperformed by a Wiener filter designed in an alternate wavelet domain. Our analysis furthermore provides a method for selecting the various parameters involved in a wavelet-domain Wiener filtering scheme.

Distributed camera network localization

Localization, estimating the positions and orientations of a set of cameras, is a critical first step in camera-based sensor network applications such as geometric estimation, scene reconstruction, and motion tracking. We propose a new distributed localization algorithm for networks of cameras with sparse overlapping view structure that is energy efficient and copes well with networking dynamics. The distributed nature of the localization computations can result in order-of-magnitude savings in communication energy over centralized approaches.

Distributed wavelet transform for irregular sensor network grids

Wavelet-based distributed data processing holds much promise for sensor networks; however, irregular sensor node placement precludes the direct application of standard wavelet techniques. In this paper, we develop a new distributed wavelet transform based on lifting that takes into account irregular sampling and provides a piecewise-planar multiresolution representation of the sensed data. We develop the transform theory; outline how to implement it in a multi-hop, wireless sensor network; and illustrate with several simulations. The new transform performs on par with conventional wavelet methods in a head-to-head comparison on a regular grid of sensor nodes

Wavelet-based deconvolution for ill-conditioned systems

In this paper, we propose a new approach to wavelet-based deconvolution. Roughly speaking, the algorithm comprises Fourier-domain system inversion followed by wavelet-domain noise suppression. Our approach subsumes a number of other wavelet-based deconvolution methods. In contrast to other wavelet-based approaches, however, we employ a regularized inverse filter, which allows the algorithm to operate even when the inverse system is ill-conditioned or non-invertible. Using a mean-square-error metric, we strike an optimal balance between Fourier-domain and wavelet-domain regularization. The result is a fast deconvolution algorithm ideally suited to signals and images with edges and other singularities. In simulations with real data, the algorithm outperforms the LTI Wiener filter and other wavelet-based deconvolution algorithms in terms of both visual quality and MSE performance.

The multiscale structure of non-differentiable image manifolds

In this paper, we study families of images generated by varying a parameter that controls the appearance of the object/scene in each image. Each image is viewed as a point in high-dimensional space; the family of images forms a low-dimensional submanifold that we call an image appearance manifold (IAM). We conduct a detailed study of some representative IAMs generated by translations/rotations of simple objects in the plane and by rotations of objects in 3-D space. Our central, somewhat surprising, finding is that IAMs generated by images with sharp edges are nowhere differentiable. Moreover, IAMs have an inherent multiscale structure in that approximate tangent planes fitted to ε-neighborhoods continually twist off into new dimensions as the scale parameter ε varies. We explore and explain this phenomenon. An additional, more exotic kind of local non-differentiability happens at some exceptional parameter points where occlusions cause image edges to disappear. These non-differentiabilities help to understand some key phenomena in image processing. They imply that Newtons method will not work in general for image registration, but that a multiscale Newtons method will work. Such a multiscale Newtons method is similar to existing coarse-to-fine differential estimation algorithms for image registration; the manifold perspective offers a well-founded theoretical motivation for the multiscale approach and allows quantitative study of convergence and approximation. The manifold viewpoint is also generalizable to other image understanding problems.

Multiple wavelet basis image denoising using Besov ball projections

We propose a new image denoising algorithm that exploits an images representation in multiple wavelet domains. Besov balls are convex sets of images whose Besov norms are bounded from above by their radii. Projecting an image onto a Besov ball of proper radius corresponds to a type of wavelet shrinkage for image denoising. By defining Besov balls in multiple wavelet domains and projecting onto their intersection using the projection onto convex sets (POCS) algorithm, we obtain an estimate that effectively combines estimates from multiple wavelet domains. While simple, the algorithm provides significant improvement over conventional wavelet shrinkage algorithms based on a single wavelet domain.