Nitin Singhal

Design and Performance Evaluation of Image Processing Algorithms on GPUs

In this paper, we construe key factors in design and evaluation of image processing algorithms on the massive parallel graphics processing units (GPUs) using the compute unified device architecture (CUDA) programming model. A set of metrics, customized for image processing, is proposed to quantitatively evaluate algorithm characteristics. In addition, we show that a range of image processing algorithms map readily to CUDA using multiview stereo matching, linear feature extraction, JPEG2000 image encoding, and nonphotorealistic rendering (NPR) as our example applications. The algorithms are carefully selected from major domains of image processing, so they inherently contain a variety of subalgorithms with diverse characteristics when implemented on the GPU. Performance is evaluated in terms of execution time and is compared to the fastest host-only version implemented using OpenMP. It is shown that the observed speedup varies extensively depending on the characteristics of each algorithm. Intensive analysis is conducted to show the appropriateness of the proposed metrics in predicting the effectiveness of an application for parallel implementation.

Robust image watermarking using local Zernike moments

In this work, we propose a robust image watermarking algorithm using local Zernike moments, which are computed over circular patches around feature points. The proposed algorithm locally computes Zernike moments and modifies them to embed watermarks, achieving robustness against cropping and local geometric attacks. Moreover, to deal with scaling attacks, the proposed algorithm extracts salient region parameters, which consist of an invariant centroid and a salient scale, and transmits them to the decoder. The parameters are used at the decoder to normalize a suspect image and detect watermarks. Extensive simulation results show that the proposed algorithm detects watermarks with low error rates, even if watermarked images are distorted by various geometric attacks as well as signal processing attacks.

Implementation and optimization of image processing algorithms on handheld GPU

The advent of GPUs with programmable shaders on handheld devices has motivated embedded application developers to utilize GPU to offload computationally intensive tasks and relieve the burden from embedded CPU. In this work, we propose an image processing toolkit on handheld GPU with programmable shaders using OpenGL ES 2.0 API. By using the image processing toolkit, we show that a range of image processing algorithms map readily to handheld GPU. We employ real-time video scaling, cartoon-style non-photorealistic rendering, and Harris corner detector as our example applications. In addition, we propose techniques to achieve increased performance with optimized shader design and efficient sharing of GPU workload between vertex and fragment shaders. Performance is evaluated in terms of frames per second at varying video stream resolution.

Hybrid approach for automatic segmentation of fetal abdomen from ultrasound images using deep learning

In this paper, we propose a hybrid approach combining traditional texture analysis methods with deep learning for the automatic detection and measurement of abdominal contour from 2-D fetal ultrasound images. Following a learning-based procedure for region of interest (ROI) localization to segment the abdominal boundary, we show that convolutional neural networks (CNNs) outperform other state-of-the-art texture features and conventional classifiers, in addressing the binary classification problem of distinguishing between abdomen versus non-abdomen regions. However, we obtain significantly better segmentation results in identifying the best ROI containing fetal abdomen, when the predictions from CNN are combined with those from gradient boosting machine (GBM) using histogram of oriented gradient (HOG) features. We trained our method on a set of 70 images and tested them on another distinct set of 70 images. We obtained a mean DICE similarity coefficient of 0.90, which shows excellent overlap with the ground truth. We report that the mean computed gestational age difference between our segmentation results and the ground truth, is within two weeks for 90% (and within one week for 70%) of the testing cases.

Object oriented framework for real-time image processing on GPU

In this paper, we present a framework for efficiently integrating programming resources of both GPU and CPU. We introduce an object oriented framework for GPGPU-based image processing. We illustrate a set of classes exploiting the design and programming advantages of an object oriented language, such as code reusability/extensibility, flexibility, information hiding, and complexity hiding. This class structure is supplemented with shader (GLSL) and kernel (CUDA) programming to facilitate full functionality. We demonstrate the potential of our approach with application scenarios and discuss the frameworks performance in terms of programming effort, execution overhead, and speedup factor achieved over CPU.

Design and optimization of image processing algorithms on mobile GPU

The advent of GPUs with programmable shaders on mobile phones has motivated developers to utilize GPU to offload computationally intensive tasks and relive the burden of embedded CPU. In this paper, we present a set of metrics to measure characteristics of a mobile phone GPU with the focus on image processing algorithms. These measures assist users in design and implementation stage and in classifying bottlenecks. We propose techniques to achieve increased performance with optimized shader design. To show the effectiveness of the proposed techniques, we employ cartoon-style non-photorealistic rendering (NPR), belief propagation (BP) stereo matching [Yang et al. 2006], and speeded up robust features (SURF) detection [Bay et al. 2008] as our example algorithms.

Mobile photo collage

In this paper, we propose an efficient technique for creating a visually appealing collage on a mobile platform from a set of input images. The proposed algorithm consists of four main modules, namely image ranking, region of interest (ROI) selection, packing, and blending. Each of the four modules is designed using a greedy and localized approach. The modules are further optimized during implementation for efficient porting on a mobile phone processor. Experimental results show the effectiveness of the proposed algorithm with visually appealing results on an off-the-shelf mobile phone.

Efficient design and implementation of visual computing algorithms on the GPU

In this paper, we explore the key factors in the design and implementation of visual computing (image processing and computer vision) algorithms on the massive parallel GPU (graphics processing units). The goal of the exploration is to provide common perspective and guidelines of using GPU for visual computing applications. We have selected three nontrivial applications (multiview stereo matching, linear feature extraction, and JPEG2000 image encoding) for the benchmarks, which show different characteristics in GPU parallel computing. Intensive analysis is performed to evaluate the characteristic of each algorithm and its effect on the performance. Based on this, we draw general guidelines of using GPU for the visual computing algorithms.

Automated assessment of endometrium from transvaginal ultrasound using Deep Learned Snake

Endometrium assessment via thickness measurement is commonly performed in routine gynecological ultrasound examination for assessing the reproductive health of patients undergoing fertility related treatments and endometrium cancer screening in women with post-menopausal bleeding. This paper introduces a fully automated technique for endometrium thickness measurement from three-dimensional transvaginal ultrasound (TVUS) images. The algorithm combines the robustness of deep neural networks with the more interpretable level set method for segmentation. We propose a hybrid variational curve propagation model which embeds a deep-learned endometrium probability map in the segmentation energy functional. This solution provides approximately 30% performance improvement over a contemporary supervised learning method on a database of 59 TVUS images and the thickness measurement is found to be within ±2mm of the manual measurement in 87% of the cases.