Several problems in imaging acquire multiple measurement vectors(MMVs)of Fourier samples for the same underlying scene.Image recovery techniques from MMVs aim to exploit the joint sparsity across the measurements in t...Several problems in imaging acquire multiple measurement vectors(MMVs)of Fourier samples for the same underlying scene.Image recovery techniques from MMVs aim to exploit the joint sparsity across the measurements in the sparse domain.This is typically accomplished by extending the use of`1 regularization of the sparse domain in the single measurement vector(SMV)case to using`2,1 regularization so that the“jointness”can be accounted for.Although effective,the approach is inherently coupled and therefore computationally inefficient.The method also does not consider current approaches in the SMV case that use spatially varying weighted`1 regularization term.The recently introduced variance based joint sparsity(VBJS)recovery method uses the variance across the measurements in the sparse domain to produce a weighted MMV method that is more accurate and more efficient than the standard`2,1 approach.The efficiency is due to the decoupling of the measurement vectors,with the increased accuracy resulting from the spatially varying weight.Motivated by these results,this paper introduces a new technique to even further reduce computational cost by eliminating the requirement to first approximate the underlying image in order to construct the weights.Eliminating this preprocessing step moreover reduces the amount of information lost from the data,so that our method is more accurate.Numerical examples provided in the paper verify these benefits.展开更多
This paper presents an application of the sparse Bayesian learning(SBL)algorithm to linear inverse problems with a high order total variation(HOTV)sparsity prior.For the problem of sparse signal recovery,SBL often pro...This paper presents an application of the sparse Bayesian learning(SBL)algorithm to linear inverse problems with a high order total variation(HOTV)sparsity prior.For the problem of sparse signal recovery,SBL often produces more accurate estimates than maximum a posteriori estimates,including those that useℓ1 regularization.Moreover,rather than a single signal estimate,SBL yields a full posterior density estimate which can be used for uncertainty quantification.However,SBL is only immediately applicable to problems having a direct sparsity prior,or to those that can be formed via synthesis.This paper demonstrates how a problem with an HOTV sparsity prior can be formulated via synthesis,and then utilizes SBL.This expands the class of problems available to Bayesian learning to include,e.g.,inverse problems dealing with the recovery of piecewise smooth functions or signals from data.Numerical examples are provided to demonstrate how this new technique is effectively employed.展开更多
Detection of edges in piecewise smooth functions is important in many applications.Higher order reconstruction algorithms in image processing and post processing of numerical solutions to partial differential equation...Detection of edges in piecewise smooth functions is important in many applications.Higher order reconstruction algorithms in image processing and post processing of numerical solutions to partial differential equations require the identification of smooth domains,creating the need for algorithms that will accurately identify discontinuities in a given function as well as those in its gradient.This work expands the use of the polynomial annihilation edge detector,(Archibald,Gelb and Yoon,2005),to locate discontinuities in the gradient given irregularly sampled point values of a continuous function.The idea is to preprocess the given data by calculating the derivative,and then to use the polynomial annihilation edge detector to locate the jumps in the derivative.We compare our results to other recently developed methods.展开更多
We investigate numerical approximations based on polynomials that are orthogonal with respect to a weighted discrete inner product and develop an algorithm for solving time dependent differential equations.We focus on...We investigate numerical approximations based on polynomials that are orthogonal with respect to a weighted discrete inner product and develop an algorithm for solving time dependent differential equations.We focus on the family of super Gaussian weight functions and derive a criterion for the choice of parameters that provides good accuracy and stability for the time evolution of partial differential equations.Our results show that this approach circumvents the problems related to the Runge phenomenon on equally spaced nodes and provides high accuracy in space.For time stability,small corrections near the ends of the interval are computed using local polynomial interpolation.Several numerical experiments illustrate the performance of the method.展开更多
This paper designs a segmentation method for an image based on its Fourier spectral data.An edge map is generated directly from the Fourier coefficients without first reconstructing the image in pixelated form.Consequ...This paper designs a segmentation method for an image based on its Fourier spectral data.An edge map is generated directly from the Fourier coefficients without first reconstructing the image in pixelated form.Consequently the internal boundaries of the edge map are not blurred by any(filtered)Fourier reconstruction.The edge map is then processed with an edge linking segmentation algorithm.We include examples from magnetic resonance imaging(MRI).Our results illustrate some potential benefits of using high order methods in imaging.展开更多
文摘Several problems in imaging acquire multiple measurement vectors(MMVs)of Fourier samples for the same underlying scene.Image recovery techniques from MMVs aim to exploit the joint sparsity across the measurements in the sparse domain.This is typically accomplished by extending the use of`1 regularization of the sparse domain in the single measurement vector(SMV)case to using`2,1 regularization so that the“jointness”can be accounted for.Although effective,the approach is inherently coupled and therefore computationally inefficient.The method also does not consider current approaches in the SMV case that use spatially varying weighted`1 regularization term.The recently introduced variance based joint sparsity(VBJS)recovery method uses the variance across the measurements in the sparse domain to produce a weighted MMV method that is more accurate and more efficient than the standard`2,1 approach.The efficiency is due to the decoupling of the measurement vectors,with the increased accuracy resulting from the spatially varying weight.Motivated by these results,this paper introduces a new technique to even further reduce computational cost by eliminating the requirement to first approximate the underlying image in order to construct the weights.Eliminating this preprocessing step moreover reduces the amount of information lost from the data,so that our method is more accurate.Numerical examples provided in the paper verify these benefits.
基金supported in part by NSF-DMS 1502640,NSF-DMS 1912685,AFOSR FA9550-18-1-0316Office of Naval Research MURI grant N00014-20-1-2595.
文摘This paper presents an application of the sparse Bayesian learning(SBL)algorithm to linear inverse problems with a high order total variation(HOTV)sparsity prior.For the problem of sparse signal recovery,SBL often produces more accurate estimates than maximum a posteriori estimates,including those that useℓ1 regularization.Moreover,rather than a single signal estimate,SBL yields a full posterior density estimate which can be used for uncertainty quantification.However,SBL is only immediately applicable to problems having a direct sparsity prior,or to those that can be formed via synthesis.This paper demonstrates how a problem with an HOTV sparsity prior can be formulated via synthesis,and then utilizes SBL.This expands the class of problems available to Bayesian learning to include,e.g.,inverse problems dealing with the recovery of piecewise smooth functions or signals from data.Numerical examples are provided to demonstrate how this new technique is effectively employed.
基金This work was partially supported by NSF grants CNS 0324957DMS 0617867+2 种基金DMS 0608844(AG)DMS 0510813(AG and HM)DMS 0421846(AG and HM).
文摘Detection of edges in piecewise smooth functions is important in many applications.Higher order reconstruction algorithms in image processing and post processing of numerical solutions to partial differential equations require the identification of smooth domains,creating the need for algorithms that will accurately identify discontinuities in a given function as well as those in its gradient.This work expands the use of the polynomial annihilation edge detector,(Archibald,Gelb and Yoon,2005),to locate discontinuities in the gradient given irregularly sampled point values of a continuous function.The idea is to preprocess the given data by calculating the derivative,and then to use the polynomial annihilation edge detector to locate the jumps in the derivative.We compare our results to other recently developed methods.
文摘We investigate numerical approximations based on polynomials that are orthogonal with respect to a weighted discrete inner product and develop an algorithm for solving time dependent differential equations.We focus on the family of super Gaussian weight functions and derive a criterion for the choice of parameters that provides good accuracy and stability for the time evolution of partial differential equations.Our results show that this approach circumvents the problems related to the Runge phenomenon on equally spaced nodes and provides high accuracy in space.For time stability,small corrections near the ends of the interval are computed using local polynomial interpolation.Several numerical experiments illustrate the performance of the method.
基金This work was partially supported by NSF grants CNS 0324957,DMS 0510813DMS 0652833NIH grant EB 02553301(AG)The first author would also like to thank the ICOSAHOM committee for the invitation to speak at this conference.
文摘This paper designs a segmentation method for an image based on its Fourier spectral data.An edge map is generated directly from the Fourier coefficients without first reconstructing the image in pixelated form.Consequently the internal boundaries of the edge map are not blurred by any(filtered)Fourier reconstruction.The edge map is then processed with an edge linking segmentation algorithm.We include examples from magnetic resonance imaging(MRI).Our results illustrate some potential benefits of using high order methods in imaging.
