The Experts below are selected from a list of 843 Experts worldwide ranked by ideXlab platform
Leonid P. Yaroslavsky - One of the best experts on this subject based on the ideXlab platform.
-
Real-Time Image and Video Processing - How fast can one arbitrarily and precisely scale images
Real-Time Image and Video Processing 2013, 2013Co-Authors: Leonid Bilevich, Leonid P. YaroslavskyAbstract:Image scaling is a frequent operation in video processing for optical metrology. In the paper, results of comparative study of computational complexity of different algorithms for scaling digital images with arbitrary scaling factors are presented and discussed. The following algorithms were compared: different types of spatial domain processing algorithms (linear, cubic, cubic spline Interpolation) and a new DCT-based algorithm, which implements perfect (Interpolation error free) scaling through discrete Sinc-Interpolation and is virtually free of boundary effects (characteristic for the DFT-based scaling algorithms). The comparison results enable evaluation of the feasibility of realtime implementation of the algorithms for arbitrary image scaling.
-
Image Processing: Algorithms and Systems - Fast DCT-based algorithm for signal and image accurate scaling
Image Processing: Algorithms and Systems XI, 2013Co-Authors: Leonid Bilevich, Leonid P. YaroslavskyAbstract:A new DCT-based algorithm for signal and image scaling by arbitrary factor is presented. The algorithm is virtually free of boundary effects and implements the discrete Sinc-Interpolation, which preserves the spectral content of the signal, and therefore is free from Interpolation errors. Being implemented through the fast FFT-type DCT algorithm, the scaling algorithm has computational complexity of O(log[σ N ]) operations per output sample, where N and [σ N ] are number of signal input and output samples, correspondingly.
-
Boundary effect free and adaptive discrete signal Sinc-Interpolation algorithms for signal and image resampling.
Applied optics, 2003Co-Authors: Leonid P. YaroslavskyAbstract:The problem of digital signal and image resampling with discrete Sinc Interpolation is addressed. Discrete Sinc Interpolation is theoretically the best one among the digital convolution-based signal resampling methods because it does not distort the signal as defined by its samples and is completely reversible. However, Sinc Interpolation is frequently not considered in applications because it suffers from boundary effects, tends to produce signal oscillations at the image edges, and has relatively high computational complexity when irregular signal resampling is required. A solution that enables the elimination of these limitations of the discrete Sinc Interpolation is suggested. Two flexible and computationally efficient algorithms for boundary effects free and adaptive discrete Sinc Interpolation are presented: frame-wise (global) Sinc Interpolation in the discrete cosine transform (DCT) domain and local adaptive Sinc Interpolation in the DCT domain of a sliding window. The latter offers options not available with other Interpolation methods: Interpolation with simultaneous signal restoration/enhancement and adaptive Interpolation with super resolution.
-
Image Processing: Algorithms and Systems - Fast signal Sinc-Interpolation methods for signal and image resampling
Image Processing: Algorithms and Systems, 2002Co-Authors: Leonid P. YaroslavskyAbstract:Digital signal resampling is required in many digital signal and image processing applications. Among the digital convolution based signal resampling methods, Sinc-Interpolation is theoretically the best one Since it does not distort the signal defined by its samples. Discrete Sinc-Interpolation is most frequently implemented by the 'signal spectrum zero padding method.' However, this method is very inefficient and inflexible. Sinc-Interpolation badly suffers also from boundary effects. In the paper, a flexible and computationally efficient methods for boundary effects free discrete Sinc-Interpolation are presented in two modifications: frame (global) Sinc-Interpolation in DCT domain and Sinc-Interpolation in sliding widow (local). In sliding window Interpolation, Interpolation kernel is a windowed Sinc-function. Windowed Sinc-Interpolation offers options not available with other Interpolation methods: Interpolation with simultaneous local adaptive signal denoising and adaptive Interpolation with super resolution. The methods outperform other existing discrete signal Interpolation methods in terms of the Interpolation accuracy and flexibility of the Interpolation kernel design. Their computational complexity is O[log(Size of the frame)] per output sample for frame Interpolation and O(Window Size) per output sample for sliding window Interpolation.
-
Fast signal Sinc-Interpolation methods for signal and image resampling
2002Co-Authors: Leonid P. YaroslavskyAbstract:Digital signal resampling is required in many digital signal and image processing applications. Among the digital convolution based signal resampling methods, Sinc-Interpolation is theoretically the best one Since it does not distort the signal defined by its samples. Discrete Sinc-Interpolation is most frequently implemented by the signal spectrum zero padding method. However, this method is very inefficient and inflexible. Sinc-Interpolation badly suffers also from boundary effects. In the paper, a flexible and computationally efficient methods for boundary effects free discrete Sinc-Interpolation are presented in two modifications: frame (global) Sinc-Interpolation in DCT domain and Sinc-Interpolation in sliding window (local). In sliding window Interpolation, Interpolation kernel is a windowed Sinc-function. Windowed Sinc-Interpolation offers options not available with other Interpolation methods: Interpolation with simultaneous local adaptive signal denoising and adaptive Interpolation with super resolution. The methods outperform other existing discrete signal Interpolation methods in terms of the Interpolation accuracy and flexibility of the Interpolation kernel design. Their computational complexity is O(log(Size of the frame)) per output sample for frame Interpolation and O(Window Size) per output sample for sliding window Interpolation.
Du Xiao-yong - One of the best experts on this subject based on the ideXlab platform.
-
A New Method for Improving SAR Resolution Based on Synthesized Bandwidth Technique
Signal Processing, 2010Co-Authors: Du Xiao-yongAbstract:Recently,a synthesized high-bandwidth technique about multi chirp pulse signal is concerned by the radar field.This technique can increase the range resolution,while reduce the instantaneous bandwidth and sample rate requirements of radar systems.Against the problem that parameters of the child pulse signal such as data sampling rate,bandwidth,and band-step may be different,we provide the method of Sinc Interpolation to reconstruct the uniform sampling signal in this paper,and fulfill the application of synthesized bandwidth technique in SAR by means of fore treatment.At last,the simulation shows that the method is applicable to the multiband child pulse signals with different parameters,and improves the range resolution effectually.
Alan V. Oppenheim - One of the best experts on this subject based on the ideXlab platform.
-
Sinc Interpolation of Nonuniform Samples
IEEE Transactions on Signal Processing, 2011Co-Authors: Shay Maymon, Alan V. OppenheimAbstract:It is well known that a bandlimited signal can be uniquely recovered from nonuniformly spaced samples under certain conditions on the nonuniform grid and provided that the average sampling rate meets or exceeds the Nyquist rate. However, reconstruction of the continuous-time signal from nonuniform samples is typically more difficult to implement than from uniform samples. Motivated by the fact that Sinc Interpolation results in perfect reconstruction for uniform sampling, we develop a class of approximate reconstruction methods from nonuniform samples based on the use of time-invariant lowpass filtering, i.e., Sinc Interpolation. The methods discussed consist of four cases incorporated in a single framework. The case of sub-Nyquist sampling is also discussed and nonuniform sampling is shown as a possible approach to mitigating the impact of aliasing.
-
randomized Sinc Interpolation of nonuniform samples
European Signal Processing Conference, 2009Co-Authors: Shay Maymon, Alan V. OppenheimAbstract:It is well known that a bandlimited signal can be uniquely determined from nonuniformly spaced samples, provided that the average sampling rate exceeds the Nyquist rate. However, reconstruction of the continuous-time signal from nonuniform samples is more difficult than from uniform samples. This paper develops and compares simpler approximate methods for signal reconstruction from nonuniform samples.
-
EUSIPCO - Randomized Sinc Interpolation of nonuniform samples
2009Co-Authors: Shay Maymon, Alan V. OppenheimAbstract:It is well known that a bandlimited signal can be uniquely determined from nonuniformly spaced samples, provided that the average sampling rate exceeds the Nyquist rate. However, reconstruction of the continuous-time signal from nonuniform samples is more difficult than from uniform samples. This paper develops and compares simpler approximate methods for signal reconstruction from nonuniform samples.
Mingyang Shang - One of the best experts on this subject based on the ideXlab platform.
-
an improved imaging algorithm for high resolution spotlight sar with continuous pri variation based on modified Sinc Interpolation
Sensors, 2019Co-Authors: Shiyang Chen, Lijia Huang, Mingyang ShangAbstract:This paper focuses on an improved imaging algorithm for spotlight synthetic aperture radar (SAR) with continuous Pulse Repetition Interval (PRI) variation in extremely high-resolution. Conventional SAR systems are limited in that a wide swath cannot be achieved with a high azimuth resolution in the meantime. This limitation can be overcome by Pulse Repetition Frequency (PRF) variation in a SAR system. However, there are problems such as the ambiguities of point targets or extended targets caused by nonuniform sampling. A reconstructive method, Nonuniform Discrete Fourier Transform (NUDFT) has been presented in the current literature, but it is rather computationally expensive. In this paper, a modified Sinc Interpolation based on NUDFT is proposed, which is used to reconstruct the uniformly sampled echo in time domain. Since the Interpolation kernel length is relatively short, it is more computationally efficient. Then, the two-step processing approach combined with the modified Sinc Interpolation is further presented, which has much better accuracy than that combined with the conventional Sinc Interpolation. Both the simulated data and the extracted GF-3 data experiment demonstrate the validity and accuracy of the proposed approach.
Saibal Mukhopadhyay - One of the best experts on this subject based on the ideXlab platform.
-
design of an energy efficient accelerator for training of convolutional neural networks using frequency domain computation
Design Automation Conference, 2017Co-Authors: Burhan Ahmad Mudassar, Saibal MukhopadhyayAbstract:Convolutional neural networks (CNNs) require high computation and memory demand for training. This paper presents the design of a frequency-domain accelerator for energy-efficient CNN training. With Fourier representations of parameters, we replace convolutions with simpler pointwise multiplications. To eliminate the Fourier transforms at every layer, we train the network entirely in the frequency domain using approximate frequency-domain nonlinear operations. We further reduce computation and memory requirements using Sinc Interpolation and Hermitian symmetry. The accelerator is designed and synthesized in 28nm CMOS, as well as prototyped in an FPGA. The simulation results show that the proposed accelerator significantly reduces training time and energy for a target recognition accuracy.
-
DAC - Design of an Energy-Efficient Accelerator for Training of Convolutional Neural Networks using Frequency-Domain Computation
Proceedings of the 54th Annual Design Automation Conference 2017, 2017Co-Authors: Burhan Ahmad Mudassar, Saibal MukhopadhyayAbstract:Convolutional neural networks (CNNs) require high computation and memory demand for training. This paper presents the design of a frequency-domain accelerator for energy-efficient CNN training. With Fourier representations of parameters, we replace convolutions with simpler pointwise multiplications. To eliminate the Fourier transforms at every layer, we train the network entirely in the frequency domain using approximate frequency-domain nonlinear operations. We further reduce computation and memory requirements using Sinc Interpolation and Hermitian symmetry. The accelerator is designed and synthesized in 28nm CMOS, as well as prototyped in an FPGA. The simulation results show that the proposed accelerator significantly reduces training time and energy for a target recognition accuracy.