Conference Paper

A fast weighted median algorithm based on Quickselect

Dept. of Electr. & Comput. Eng., Univ. of Delaware, Delaware, OH, USA
DOI: 10.1109/ICIP.2010.5651855 Conference: Image Processing (ICIP), 2010 17th IEEE International Conference on
Source: IEEE Xplore

ABSTRACT Weighted median filters are increasingly being used in signal processing applications and thus fast implementations are of importance. This paper introduces a fast algorithm to compute the weighted median of N samples which has linear time and space complexity as opposed to O(N logN) which is the time complexity of traditional sorting algorithms. The proposed algorithm is based on Quickselect which is closely related to the well known Quicksort. We compare the runtime and the complexity to Floyd and Rivest's algorithm SELECT which to date has been the fastest median finding algorithm and show that our algorithm is up to 30% faster.

0 Followers
 · 
776 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Accurate reconstruction of 3D geometrical shape from a set of calibrated 2D multiview images is an active yet challenging task in computer vision. The existing multiview stereo methods usually perform poorly in recovering deeply concave and thinly protruding structures, and suffer from several common problems like slow convergence, sensitivity to initial conditions, and high memory requirements. To address these issues, we propose a two-phase optimization method for generalized reprojection error minimization (TwGREM), where a generalized framework of reprojection error is proposed to integrate stereo and silhouette cues into a unified energy function. For the minimization of the function, we first introduce a convex relaxation on 3D volumetric grids which can be efficiently solved using variable splitting and Chambolle projection. Then, the resulting surface is parameterized as a triangle mesh and refined using surface evolution to obtain a high-quality 3D reconstruction. Our comparative experiments with several state-of-the-art methods show that the performance of TwGREM based 3D reconstruction is among the highest with respect to accuracy and efficiency, especially for data with smooth texture and sparsely sampled viewpoints.
    Image and Vision Computing 11/2014; 33. DOI:10.1016/j.imavis.2014.10.008 · 1.58 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Low-rank matrix approximation plays an important role in the area of computer vision and image processing. Most of the conventional low-rank matrix approximation methods are based on the l2 -norm (Frobenius norm) with principal component analysis (PCA) being the most popular among them. However, this can give a poor approximation for data contaminated by outliers (including missing data), because the l2 -norm exaggerates the negative effect of outliers. Recently, to overcome this problem, various methods based on the l1 -norm, such as robust PCA methods, have been proposed for low-rank matrix approximation. Despite the robustness of the methods, they require heavy computational effort and substantial memory for high-dimensional data, which is impractical for real-world problems. In this paper, we propose two efficient low-rank factorization methods based on the l1 -norm that find proper projection and coefficient matrices using the alternating rectified gradient method. The proposed methods are applied to a number of low-rank matrix approximation problems to demonstrate their efficiency and robustness. The experimental results show that our proposals are efficient in both execution time and reconstruction performance unlike other state-of-the-art methods.
    IEEE transactions on neural networks and learning systems 02/2015; 26(2):237-51. DOI:10.1109/TNNLS.2014.2312535 · 4.37 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a simple and robust algorithm for compressive sensing (CS) signal reconstruction based on the weighted median (WM) operator. The proposed approach addresses the reconstruction problem by solving a l<sub>0</sub>-regularized least absolute deviation (l<sub>0</sub>-LAD) regression problem with a tunable regularization parameter, being suitable for applications where the underlying contamination follows a statistical model with heavier-than-Gaussian tails. The solution to this regularized LAD regression problem is efficiently computed, under a coordinate descent framework, by an iterative algorithm that comprises two stages. In the first stage, an estimation of the sparse signal is found by recasting the reconstruction problem as a parameter location estimation for each entry in the sparse vector leading to the minimization of a sum of weighted absolute deviations. The solution to this one-dimensional minimization problem turns out to be the WM operator acting on a shifted-and-scaled version of the measurement samples with weights taken from the entries in the measurement matrix. The resultant estimated value is then passed to a second stage that identifies whether the corresponding entry is relevant or not. This stage is achieved by a hard threshold operator with adaptable thresholding parameter that is suitably tuned as the algorithm progresses. This two-stage operation, WM operator followed by a hard threshold operator, adds the desired robustness to the estimation of the sparse signal and, at the same time, ensures the sparsity of the solution. Extensive simulations demonstrate the reconstruction capability of the proposed approach under different noise models. We compare the performance of the proposed approach to those yielded by state-of-the-art CS reconstruction algorithms showing that our approach achieves a better performance for different noise distributions. In particular, as the distribution tails become heavier the performance gai n achieved by the proposed approach increases significantly.
    Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2010, 14-19 March 2010, Sheraton Dallas Hotel, Dallas, Texas, USA; 01/2010

Preview

Download
77 Downloads
Available from