Accelerating SIFT on parallel architectures

Warn, Seth; Emeneker, Wesley; Cothren, Jackson; Apon, Amy

doi:10.1109/clustr.2009.5289155

Cited by 36 publications

(14 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As evaluated in [24], their design actually gained a speedup of 6.0X from its parallelism for 16-core machine. Moreover, Seth et al [20] explored and evaluated parallel SIFT using large images as input in both multi-core CPU and GPGPU. They found that SIFT requires huge memory space when the input image size is large.…”

Section: B Previous Acceleration Effortsmentioning

confidence: 99%

“…However, no satisfied results are achieved in these works. For example, the researches in [15] [20] only achieve about a speedup of 6X on a 16-core machine. It is obvious that a large gap exists between their results and the ideal speedup, which means that there are some limitations in their parallel implementations or in original serial IFEAs.…”

Section: Introductionmentioning

confidence: 97%

See 1 more Smart Citation

Adaptive Pipeline Parallelism for Image Feature Extraction Algorithms

Chen

Yang

Zhang

et al. 2012

2012 41st International Conference on Parallel Processing

View full text Add to dashboard Cite

Currently, multimedia data has become one of the major data types processed and transferred on the Internet. With the rapid growth of multimedia data, it is vitally important to find an efficient way to extract useful information from a large amount of data. SIFT and SURF, as the most popular multimedia feature extraction algorithms, have been widely used in many applications. However, the limited processing speed (about 1.8 and 2.6 images or frames per second for SIFT and SURF respectively on an ordinary CPU) makes it impossible to apply them in many real-world applications with real-time requirements. Therefore, it has become one of the major challenges that how to improve the processing speed of these multimedia feature extraction algorithms.The popularity of multi-core architecture and the increase of computation resources on different platforms provide a new opportunity to accelerate the processing speed of these image feature extraction algorithms. In this paper, we first systematically analyze the major parallel constraints in SIFT and SURF, such as imbalanced workload and indeterminate time distribution. Then, based on these analysis, we design and implement an adaptive pipeline parallel scheme (AD-PIPE) for both SIFT and SURF to alleviate these limitations. In our scheme, we dynamically check the workload in different pipeline stages and adjust the thread number in different stages to achieve a balanced partition. Experimental results show that our approach is efficient and scalable. It can achieve a speedup of 16.88X and 20.33X respectively for SIFT and SURF on a 16-core machine and a realtime processing speed with about 27 and 52 images or frames per second.

show abstract

Section: B Previous Acceleration Effortsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 97%

Adaptive Pipeline Parallelism for Image Feature Extraction Algorithms

Chen

Yang

Zhang

et al. 2012

2012 41st International Conference on Parallel Processing

View full text Add to dashboard Cite

show abstract

“…And this approach has been further developed in [10], which utilizes both CPU and GPU by combining OpenMP, SSE (Streaming SIMD Extension) and CUDA (Compute Unified Device Architecture) programming. Then, [11] applied GPU and OpenMP based SIFT implementation to massive satellite image processing. One of the disadvantages of this approach is that special hardware is required, which increased the cost.…”

Section: Related Workmentioning

confidence: 99%

A Distributed Dynamic Parallel Algorithm for SIFT Feature Extraction

Jiang¹,

Zhang²,

Zhang³

2010

2010 3rd International Symposium on Parallel Architectures, Algorithms and Programming

View full text Add to dashboard Cite

This paper deals with the issue of developing efficient algorithms for accelerating SIFT (Scale Invariant Feature Transform) features extraction under distributed environment. The proposed distributed dynamic parallel algorithm (DDP-SIFT) using a special data parallel approach that divides the Gauss Scale Space by octave aimed at acquiring large image blocks which is of great importance in some application. To make this approach effective, steps of building Gauss Scale Space are changed, and only the prerequisite part which is only 1/13 of the whole pyramid will be produced before tasks allocation, and Allocation Data Quantity (ADQ) is decreased by 13 times. Data blocks are assembled as tasks maintained in task lists, and dynamically allocated to Computing Nodes. A refined-blocking approach is proposed to further improve load balance. Our investigations show that the proposed algorithm has remarkable performance on accelerating SIFT features extraction while pursuing large data blocks.

show abstract

“…GPUs have been shown to provide great speedup on a wide range of computer vision algorithms, including SIFT [7,8,9,10]. However, this success has mostly been confined to desktop GPU applications.…”

Section: Prior Workmentioning

confidence: 99%

A fast and efficient sift detector using the mobile GPU

Rister

Wang

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

View full text Add to dashboard Cite

Emerging mobile applications, such as augmented reality, demand robust feature detection at high frame rates. We present an implementation of the popular Scale-Invariant Feature Transform (SIFT) feature detection algorithm that incorporates the powerful graphics processing unit (GPU) in mobile devices. Where the usual GPU methods are inefficient on mobile hardware, we propose a heterogeneous dataflow scheme. By methodically partitioning the computation, compressing the data for memory transfers, and taking into account the unique challenges that arise out of the mobile GPU, we are able to achieve a speedup of 4-7x over an optimized CPU version, and a 6.4x speedup over a published GPU implementation. Additionally, we reduce energy consumption by 87 percent per image. We achieve near-realtime detection without compromising the original algorithm.

show abstract

Accelerating SIFT on parallel architectures

Cited by 36 publications

References 5 publications

Adaptive Pipeline Parallelism for Image Feature Extraction Algorithms

Adaptive Pipeline Parallelism for Image Feature Extraction Algorithms

A Distributed Dynamic Parallel Algorithm for SIFT Feature Extraction

A fast and efficient sift detector using the mobile GPU

Contact Info

Product

Resources

About