We study the problem of recovering point sources from samples of their convolution with a Gaussian kernel, showing that a convex program achieves exact deconvolution as long as the sources are not too clustered togeth...
详细信息
ISBN:
(纸本)9781538615669
We study the problem of recovering point sources from samples of their convolution with a Gaussian kernel, showing that a convex program achieves exact deconvolution as long as the sources are not too clustered together and there are at least two samples close to the location of each source. The result is established using a novel dual-certificate construction.
We study the question of reconstructing a sequence of {f_i, g_i}_(i=1)~s from the sum of their convolution, i.e., y = ∑_(i=1)~s f_i * g_i. This problem is closely related to both blind deconvolution and blind demixin...
详细信息
ISBN:
(纸本)9781538615669
We study the question of reconstructing a sequence of {f_i, g_i}_(i=1)~s from the sum of their convolution, i.e., y = ∑_(i=1)~s f_i * g_i. This problem is closely related to both blind deconvolution and blind demixing problem. Our goal is to find all {f_i, g_i}_(i=1)~s by jointly demixing each component f_i*g_i and performing deconvolution procedure. While the convex program is able to solve this problem effectively and robustly under certain conditions, the main obstacle towards real-time deployment is to find a provably convergent, efficient and robust algorithm. We present an efficient numerical algorithm which guarantees the exact recovery of the solutions despite its notorious non-convexity. The proposed two-step algorithm converges to the global minima linearly and is also robust to the noise. Though the derived performance bound is suboptimal in terms of the information-theoretic limit, numerical simulations show the remarkable performance even if the number of measurements is close to the degree of freedom.
The impetus of this work originated from the advent of high magnetic field magnetic resonance imaging scanners with B0 fields of 4T, 7T, and 9.4T. These ultrahigh magnetic field systems generally improve the signal to...
详细信息
The impetus of this work originated from the advent of high magnetic field magnetic resonance imaging scanners with B0 fields of 4T, 7T, and 9.4T. These ultrahigh magnetic field systems generally improve the signal to noise ratios. However, B1 field non-uniformity also occurs due to the increased RF field frequencies when wavelengths in the head become shorter than its size. As interest in multiple channel transmission line coils increases, the control of the amplitude and phase of individual coil elements is required in order to develop desired B1 field. The choice of the excitation of the coil elements may be determined by convex optimization. convex optimization is used provides results very fast, when the problem is formulated globally. In addition, convex optimization provides better signal to noise (SNR) ratio when anatomic specific regions are investigated. In this paper, simulation and experimental results are discussed at 9.4T systems based on the number of elements. The primary objective of this study is to increase the signal in a specific target region and decrease the signal and noise in the outside region termed the suppression region. The convex formulations are minimizing the maximum field point in the suppression region while keeping the center of target maximum. Based on this min-max optimization criterion, an iteration method which modifies the selection of suppression fields is also performed to produce better results. The results of the localization on FDTD human data at 9.4T are shown in Fig. 1. In these figures, the axial slices of the center of human head model provided by XFDTD are used after manipulating with MATLAB and the 16 channel head coil is excited. Figure 1 shows an improvement of the homogeneity in the suppression region when the target region is at center. In Fig. 2, received signal localizations are obtained for three different regions of interest (ROI) after using the convex optimization. Note that the selection of ROI is limite
暂无评论