Samad Wali has been working as an Assistant Professor in the Department of Mathematics at Namal University, Minawali since July 2020. He obtained his Doctor of Philosophy in Computational Mathematics from Nankai University, China in 2018 and subsequently completed a two-year postdoctoral training program at the University of Electronic Science and Technology of China. Throughout his academic career, Samad has focused his research on image processing, variational methods, and optimization, and has made notable contributions in the form of developing highly efficient computational techniques for solving nonlinear partial differential equations. Specifically, his work has been centered around image restoration and segmentation.
Doctor of Philosophy
Nankai University, China
Master of Science
( Applied Mathematics)
The Islamia University of Bahawalpur
Bachelor of Science
University of Electronic Science and Technology of China
01-Sep-2018 - 01-Apr-2020
Honours and Awards
CSC Scholarship for PhD
CSC fully-funded PhD Scholarship, Chinese Scholarship Council
An Efficient Supervised Deep Hashing Method for Image Retrieval
In recent years, searching and retrieving relevant images from large databases has become an emerging challenge for the researcher. Hashing methods that mapped raw data into a short binary code have attracted increasing attention from the researcher. Most existing hashing approaches map samples to a binary vector via a single linear projection, which restricts the flexibility of those methods and leads to optimization problems. We introduce a CNN-based hashing method that uses multiple nonlinear projections to produce additional short-bit binary code to tackle this issue. Further, an end-to-end hashing system is accomplished using a convolutional neural network.
QSM for mmWave Wireless Communications: Frequency Diverse Transmitter and
Reduced Complexity Receiver.
Quadrature spatial modulation (QSM) based millimeter wave (mmWave) communication system design using frequency diverse array (FDA) is proposed in this article. Since spatial modulation techniques (SMT) employ maximum likelihood (ML) based detector at receiver, the computational complexity increases vastly. Moreover, numerous existing SMT methods utilize phased array (PA), where only angle dependent communication is possible. In this article, we propose to use a standard FDA that exploits a small linearly increasing frequency offset across the array for range-angle dependent QSM wireless communications. That is, the in-phase and quadrature components of transmission vector utilize slightly different frequencies to deliver the information. Furthermore, we propose a suboptimal multi-stage (MS) detector in the receiver, which applies a matched filtering (i.e., bandpass filters matched to the corresponding frequency offsets) approach to decode the index bits at first step, whereas, merely, two most probable estimated indices are utilized further to decode associated data bits using a standard ML method. Adhering an improved signal to noise ratio (SNR) due to an FDA based range-angle dependent transmission, simulation and numerical results show the improved performance of the proposed design over existing SM and QSM based schemes, while MS approach reduces the receiver's computational complexity.
Transmit beamspace design for FDA–MIMO radar with alternating direction method of multipliers
Hybridization of a frequency diverse array (FDA) generated range-angle-time dependent beampattern with multiple input multiple output (MIMO) radar waveform diversity, namely, FDA-MIMO radar, provides more degrees-of-freedom to improve overall system performance. As the beampattern optimization needs an efficient method to approximate the desired beampattern and simultaneously minimize the cross-correlation sidelobes, we optimize the transmit beampattern over the beamspace matrix in this paper. That is, the transmit FDA antennas are divided into multiple overlapping subarrays and each subarray transmits an orthogonal waveform towards the target, whereas the weight vectors of all subarrays jointly form a beamspace matrix. Since the optimization problem turns to be a non-convex problem with a constant energy constraint, we propose an efficient algorithm based on alternating direction method of multipliers (ADMM) to solve it, which has a very fast convergence speed. The effectiveness and superiorities of the proposed method over existing methods are verified by extensive simulation results.
Integrating deep convolutional neural networks with marker-controlled watershed for overlapping nuclei segmentation in histopathology images
Nuclei segmentation in histopathology images plays a crucial role in the morphological quantitative analysis of tissue structure and has become a hot research topic. Though numerous efforts have been tried in this research area, the overlapping and touching nuclei segmentation remains a challenging problem. In this paper, we present a novel and effective instance segmentation method for tackling this challenge by integrating Deep Convolutional Neural Networks with Marker-controlled Watershed. Firstly, we design a novel network architecture with multiple segmentation tasks, called Deep Interval-Marker-Aware Network, for learning the foreground, marker, and interval of nuclei, simultaneously. Then the learned interval between overlapping nuclei is used to refine the foreground result of nuclei by using the logical operators. Finally, the learned marker result and the nuclei segmentation result refined by interval are transmitted into the Marker-controlled Watershed for splitting the touching nuclei. The experiments on the standard public datasets demonstrate that our method achieves a substantial improvement compared with state-of-the-art methods. Source codes are available at https://github.com/appiek/Nuclei_Segmentation_Experiments_Demo
Fast and Adaptive Boosting Techniques for Variational
Based Image Restoration.
Variational based problems are an important class of problems and have a space of improvement in image processing. Boosting techniques have been shown capable of improving many image restoration algorithms. This paper considers four fast and adaptive boosting techniques for variational based image restoration. The adaptive boosting frameworks can compute the existing image restoration algorithm iteratively. The primary idea is to get an enhanced result by using the output of the current step as a part of the input for the next step. Our techniques can boost variational based regularization models like total variation (TV) and total generalized variation (TGV). For image restoration, we used an adaptive regularization parameter selection, which produces signals with more details and preserves tiny objects. For efficient numerical optimization, we implement the alternating direction method of multipliers (ADMM) and demonstrate the effectiveness of the proposed techniques with a variety of experimental results. The simulation results show that the proposed boosting techniques achieve a better restoration performance on comparisons with TV and TGV in terms of quality metrics such as signal to noise ratio (SNR) and structure similarity (SSIM).
Proximal ADMM for Euler's Elastica Based Image Decomposition Model
This paper studies image decomposition models which involve functional related to total variation and Euler's elastica energy. Such kind of variational models with first order and higher order derivatives have been widely used in image processing to accomplish advanced tasks. However, these non-linear partial differential equations usually take high computational cost by the gradient descent method. In this paper, we propose a proximal alternating direction method of multipliers (ADMM) for total variation (TV) based Vese-Osher's decomposition model [L. A. Vese and S. J. Osher, J. Sci. Comput., 19.1 (2003), pp. 553-572] and its extension with Euler's elastica regularization. We demonstrate that efficient and effective solutions to these minimization problems can be obtained by proximal based numerical algorithms. In numerical experiments, we present numerous results on image decomposition and image denoising, which conforms significant improvement of the proposed models over standard models.
An Efficient Method for Euler’s Elastica Based Image Deconvolution
Variational models involving Euler's elastica energy have a wide range of applications in digital image processing. Recently, fast methods, such as the proximal-augmented Lagrangian method (PALM), have been successfully used to solve nonlinear higher order models for image restoration. In this paper, we extend fast method PALM to Euler's elastica deconvolution models with quadratic and nonquadratic fidelity terms. The proposed variational model can eliminate blur and noise and preserve edges while reducing the blocky and staircase artifacts during image restoration. We present an efficient and effective solution to the proposed minimization problems by a proximal-based numerical scheme. Our numerical experiments demonstrate several results on image deblurring and denoising, which shows a clear improvement of the proposed model over standard variational models such as total variation and Hessian-based model.
A New Adaptive Total
Generalized Variation(TGV) Boosting Technique for Image Denoising and Inpainting.
Journal of Visual Communication and Image Representation
In this paper we present a new adaptive boosting technique for total generalized variation (TGV) based image denoising and inpainting. Instead of the strengthening and substracting steps in existing boosting techniques, the proposed technique is iteratively operated by two steps: the first step is to take average of restored image with observed image, and updated parameter; the second step is to operate the TGV restoration algorithm with the average and dynamic parameter. For each iteration, as the input contains more correct information, the restoration algorithm can produce signals with more details. We have solved our boosting TGV model by primal-dual method, and applied the boosting TGV technique for gray/color image denoising and inpainting.
A Boosting Procedure for Variational-Based Image Restoration
Variational methods are an important class of methods for general image restoration. Boosting technique has been shown capable of improving many image denoising algorithms. This paper discusses a boosting technique for general variational image restoration methods. It broadens the applications of boosting techniques to a wide range of image restoration problems, including not only denoising but also deblurring and inpainting. In particular, we combine the recent SOS technique with dynamic parameter to variational methods. The dynamic regularization parameter is motivated by Meyer's analysis on the ROF model. In each iteration of the boosting scheme, the variational model is solved by augmented Lagrangian method.
- Introduction to Linear Programming and Optimization
- Real Anaysis I
- Probability and Statistics I
- Calculus I, II