We present a primal-dual interior point method (IPM) with a novel preconditioner to solve the l(1)-norm regularized least square problem for nonnegative sparse signal reconstruction. IPM is a second-order method that ...
详细信息
ISBN:
(纸本)9781479970612
We present a primal-dual interior point method (IPM) with a novel preconditioner to solve the l(1)-norm regularized least square problem for nonnegative sparse signal reconstruction. IPM is a second-order method that uses both gradient and Hessian information to compute effective search directions and achieve super-linear convergence rates. It therefore requires many fewer iterations than first-order methods such as iterative shrinkage/thresholding algorithms (ISTA) that only achieve sub-linear convergence rates. However, each iteration of IPM is more expensive than in ISTA because it needs to evaluate an inverse of a Hessian matrix to compute the Newton direction. We propose to approximate each Hessian matrix by a diagonal matrix plus a rank-one matrix. This approximation matrix is easily invertible using the Sherman-Morrison formula, and is used as a novel preconditioner in a preconditioned conjugate gradient method to compute a truncated Newton direction. We demonstrate the efficiency of our algorithm in compressive 3D volumetric image reconstruction. Numerical experiments show favorable results of our method in comparison with previous interior point based and iterative shrinkage/thresholding based algorithms.
During transmission and reception the images are degraded by noise. Also images captured using different low quality devices adversely affect the visual quality of images. The presence of the noise results in loss of ...
详细信息
ISBN:
(纸本)9781538677094
During transmission and reception the images are degraded by noise. Also images captured using different low quality devices adversely affect the visual quality of images. The presence of the noise results in loss of visibility, gives a mottled, grainy, textured, or snowy appearance. Thus making the imagevisually unpleasing which drastically affects the human vision. In addition during imageprocessing, presence of noise makes it hard for further processing (impulse, Gaussian white, salt and pepper, adversarial etc.,). Existing methods use conventional filters and Neural network models for image denoising where they compromise with the visibility of image after rigorous iterations of denoising algorithms. In this paper we implement CGANs for image denoising and evaluate the performance of CGAN with different Neural network models viz., CNN,GAN for single or multiple image denoising problem. The qualitative performance of de-noised image/images is measured using PSNR and confusion matrix.
imageprocessing and computer vision algorithms are very computationally intensive and could not be implemented on low power microcontroller used in embedded and small IoT devices. Performing computation on the cloud ...
详细信息
ISBN:
(纸本)9781728123264
imageprocessing and computer vision algorithms are very computationally intensive and could not be implemented on low power microcontroller used in embedded and small IoT devices. Performing computation on the cloud is not practical for always-on real-time operations and deploying a high-end microcontroller or a microprocessor consume too much power and are generally too expensive to fit into these small systems. In this paper, we propose a new programming workflow using reusable hardware modules and a graphical programming interface for implementing a complete imageprocessing system on an FPGAs which overcomes the steep learning curve of tradition FPGAs design tool. The design can be deployed onto the low-ended FPGAs in a similar price point as a mid-range microcontroller. To demonstrate our proposed framework, we have implemented a number of image transformation operations on an FPGAs development board with Lattice iCE40 Ultra Plus FPGA and a tiny camera module. Results have shown that our designs can fit in a low-ended FPGAs while performing 5.5-345x faster and consuming 2.4-4x less power compared to current state-of-the-art microcontroller used in small embedded and IoT devices.
In this paper, the traffic sign recognition module of a small-scale autonomous car prototype will be presented. The process undergoing the choice of an appropriate algorithm, as well as the factors taken into consider...
详细信息
ISBN:
(数字)9781728108780
ISBN:
(纸本)9781728108797
In this paper, the traffic sign recognition module of a small-scale autonomous car prototype will be presented. The process undergoing the choice of an appropriate algorithm, as well as the factors taken into consideration will be presented in the form of a case study. The current literature presents various ways of achieving the recognition of traffic signs, but most of them are computational expensive, or have difficulty in offering consistent results in conditions that are different from the prerecorded ones. Since the processing on the car is carried on an embedded platform from Nvidia (Jetson TX2), this study is based on the same board, the stream being captured with a low-cost webcam. Classical algorithms like SURF (Speeded Up Robust Features), SIFT (Scale Invariant Feature Transform) or ORB (Oriented fast and Rotated Brief) offer reliable results when the lighting condition between the reference image and the image obtained from the camera are similar. In our setup, the algorithms mentioned above start to behave badly in low light conditions. Therefore, this paper discusses the possibility of using Haar like features alongside a classifier for detecting traffic signs.
We review the recent progress on the application of imageprocessing techniques to optical communication systems. The focus is placed mainly on the implementation complexity and performance of the techniques for optic...
详细信息
ISBN:
(纸本)9781538679364
We review the recent progress on the application of imageprocessing techniques to optical communication systems. The focus is placed mainly on the implementation complexity and performance of the techniques for optical performance monitoring and the compensation of common phase error. We also briefly introduce several applications where machine learning algorithms could be beneficial to fiber-optic transmission system.
X-ray fluoroscopy is commonly used during liver embolization procedures to guide intravascular devices (e.g. guidewire and catheter) to the branches of the hepatic arteries feeding tumors. A vascular roadmap can be cr...
详细信息
ISBN:
(纸本)9781510625440
X-ray fluoroscopy is commonly used during liver embolization procedures to guide intravascular devices (e.g. guidewire and catheter) to the branches of the hepatic arteries feeding tumors. A vascular roadmap can be created to provide a reference for the device position. Recently, techniques have been developed to create dynamic vessel masks to compensate for respiratory motion. In order to superimpose the intravascular guidewire onto the vessel mask, robust segmentation is required. Commonly used techniques often use mask subtraction to isolate the device in x-ray images. However, this is not suitable due to the motion in liver applications. The proposed method uses a deep convolutional neural network to segment the guidewire in native (unsubtracted) x-ray images. The neural network uses an encoder / decoder structure, which is based on the VGG-16 network. To create a large dataset of annotated images, simulated images were created based on 3D digital subtraction angiography acquisitions of hepatic arteries in porcine studies. Random guidewire shapes were generated within the vascular volume and superimposed on the original non-contrast projection images. The network was trained using a set of 56,768 images created from 10 acquisitions. The segmentation results of the trained network were compared to a mask-subtraction-based algorithm for an independent validation data set. The deep learning algorithm (Dice = 58.1%, false negative rate (FNR) = 9.6%) outperformed the subtraction technique (Dice = 23.7%, FNR = 40.8%). This study shows that the deep learning approach is suitable for robust segmentation of curvilinear structures such as guidewires and could be used to superimpose the segmented device on dynamic roadmaps.
Automated driving technology involves various modern in-vehicle systems that are designed to increase road traffic safety by helping drivers gain a better awareness of the road and its potential hazards as well as oth...
详细信息
ISBN:
(纸本)9781538679012;9781538679265
Automated driving technology involves various modern in-vehicle systems that are designed to increase road traffic safety by helping drivers gain a better awareness of the road and its potential hazards as well as other drivers around them. Stop sign detection module is an integral part of these systems as it has great utility in automated or assisted driving applications. Although promising results have been achieved in the areas of stop sign detection and classification, these methods are heavily dependent on images and imageprocessingalgorithms and detection problem in the real world remains a challenging issue. In this paper, we propose a method to detect the stop sign, based on statistical analysis of data obtained by drive history, a project conducted by Ford Motor Company. Our detection algorithm is based on the speed profile where speed is zero or close to zero. Then, we apply clustering to our data set to extract the points with this common feature. In the end, results demonstrate that our algorithm can improve the stop sign detection efficiency and afford high precision where other algorithms are prone to failure.
The segmentation task refers to the preliminary stage of image preprocessing. Further object detection, feature recognition, scene analysis and prediction of the situations depends on its results. Modern segmentation ...
详细信息
The segmentation task refers to the preliminary stage of image preprocessing. Further object detection, feature recognition, scene analysis and prediction of the situations depends on its results. Modern segmentation algorithms require: the refusal to use of a priori information, the availability of a quality functional for the result assessment, the variable number of segments in the partition of the original image, the linear computational complexity and the adequacy of the segmentation results. Among the methods of the cluster analysis that satisfy most of the requirements listed, the Ward's method is appropriate. But, its high computational complexity prevents its direct application. The purpose of this study is to overcome the excessively high computational complexity of the classical Ward's method. The classical methods of cluster analysis that meet the actual requirements for modern image segmentation algorithms are compared. The choice of the classical Ward's method is justified as well as its advantages and disadvantages are presented. The application of the idea of the reversible computing in imageprocessing is described. The modifications of the computational process of the Ward's method are described. A model flowchart of a sequence of algorithms allowing to bypass the problem of the computational complexity peculiar to Ward's method is proposed. The experimental results of the quality improvement of the conventional segmentation are presented. The proposed model flowchart allows one to bypath the problem of the computational complexity by dividing the process into separate sequential three stages. The model flowchart is suitable for the quality improvement of any traditional segmentation.
image mosaic could generate wide-angle images by stitching regional images with their overlapped portions, which is widely used technology in image signal processing. During the procedure, SIFT-feature based approache...
详细信息
ISBN:
(纸本)9781538673928
image mosaic could generate wide-angle images by stitching regional images with their overlapped portions, which is widely used technology in image signal processing. During the procedure, SIFT-feature based approaches are proven to perform best in such methods. However, traditional implementations of SIFT algorithm would consume the most of the processing time and energy. By balancing the processing quality and the design specifications of such applications of image mosaic, approximate computing hardware implementations have been adopted in this work to improve the energy-efficiency with ignorable quality degradations. Moreover, the approximate adders of LOA have been used to replace the exact adder to carry out the Gaussian convolution process of the algorithms, which further boost the SIFT operation performance, so-called ApproxSIFT. Simulation results of logic level show that, compared with the traditional method of exact computing, the ApproxSIFT hardware implementation with approximate adders significantly improve the image mosaic system, which achieves 2.4X and 3.5X in speed and power consumption, respectively.
暂无评论