Edge detection is a crucial step in medical imaging and in a no. of other image processing applications, such as face-identification or recognition, and other classification problems. Various methods have been developed for edge detection based on applications and edge types. Some of the most common techniques used are Sobel, Prewitt, Robert, LoG and Canny etc. However, most of these methods for edge detection of various images (including x-rays image) is a computationally expensive process in terms of both time and space. Because of this delay the patients and the doctors do not get instant information or imaging reports (for example regarding fractured bone in case of x-rays). This ultimately leads to delayed diagnosis and treatment of the patient. In this work we present our findings of research related to an important edge detection technique which involve finding image gradient. We emphasize that our approach is equally valid for many different kinds of edges in an image and not just for fractured bone. To eliminate latency issue we used a graphical processor with CUDA API to implement an image gradient. The graphical processors are massively parallel processors that come inside a graphics card and have become a standard piece of hardware on all modern day computing systems including portable hand-held device. We emphasize that alternate solutions such as FPGA (Field Programmable Gate Array) and ASIC (Application Specific Integrated Circuit) based solutions are much costlier and take much longer time for development as compared to a graphical processor which is programmable using C-CUDA. We compared our implementation’s performance with respect to a CPU-only implementation. To prove our idea we used an algorithm which is a parallel version of naïve serial algorithm. Thanks to GPU’s enormous amount of computational units, our GPU-implementation shows several fold speed ups with respect to a standard CPU-only implementation. Our proof-of-concept (PoC) developed as part of this research, thus establish that the GPU stands a very good candidate for such edge detection problems where we need faster results, i.e. in real time or in near real-time.

Digital Image, Edge detection, GPU, CUDA, X-ray, gradient

[1] Marr and E. Hildrith, “Theory of Edge Detection,” Proc. Royal Society of London, B207, pp. 187–217, 1980.
[2] James Clerk Maxwell, DIGITAL IMAGE PROCESSING Mathematical and Computational Methods.
[3] R .Gonzalez and R. Woods, Digital Image Processing, ,Addison Wesley, 1992, pp 414 - 428.
[4] S. Sridhar, Oxford university publication. , Digital Image Processing.
[5] Shamik Tiwari , Danpat Rai & co.(P) LTD. “Digital Image processing”
[6] J. F. Canny. “A computational approach to edge detection”. IEEE Trans. Pattern Anal. Machine Intell., vol.PAMI-8, no. 6, pp. 679-697, 1986 Journal of Image Processing (IJIP), Volume (3) : Issue (1)
[7] Geng Xing, Chen ken , Hu Xiaoguang “An improved Canny edge detection algorithm for color image” IEEE TRANSATION ,2012 978-1-4673-0311-8/12/$31.00 ©2012 IEEE.
[8] Punarselvam, E., & Suresh, P. (2011). Edge Detection of CT scan Spine disc image using Canny Edge Detection Algorithm based on Magnitude and Edge Length. 3rd International Conference on Trendz in Information Sciences & Computing (TISC2011). doi:10.1109/tisc.2011.6169100
[9] Nikolic, M., Tuba, E., & Tuba, M. (2016). Edge detection in medical ultrasound images using adjusted Canny edge detection algorithm. 2016 24th Telecommunications Forum (TELFOR). doi:10.1109/telfor.2016.7818878
[10] Chang, C., & Kehtarnavaz, N. (2015). Computationally efficient image deblurring using low rank image approximation and its GPU implementation. Journal of Real-Time Image Processing, 12(3), 567-573. doi:10.1007/s11554-015-0539-x
[11] Sher Jung, Rajendra Kumar Sharma, GSZRP: Graphics-hardware based Optimized Secure Zone Routing protocol, IJARSE (ISSN: 2319-8354),Volume No.06, Issue No. 12, December 2017
[12] Agarwal, N., Nellans, D., Ebrahimi, E., Wenisch, T. F., Danskin, J., & Keckler, S. W. (2016). Selective GPU caches to eliminate CPU-GPU HW cache coherence. 2016 IEEE International Symposium on High Performance Computer Architecture (HPCA). doi:10.1109/hpca.2016.7446089
[13] “Convolution,” Wikipedia, 20-May-2018. [Online]. Available: http://en.wikipedia.org/wiki/Convolution. [Accessed: 23-May-2018].
[14] R. Farber, “CUDA, Supercomputing for the Masses: Part 1,” Dr. Dobb`s. [Online]. Available: http://www.drdobbs.com/high-performance-computing/207200659. [Accessed: 23-May-2018].
[15] “Weather, Atmospheric, Ocean Modeling, and Space Sciences,” NVIDIA. [Online]. Available: http://www.nvidia.com/object/weather.html. [Accessed: 23-May-2018].
[16] “The AI Computing Company | NVIDIA.” [Online]. Available: https://www.bing.com/cr?IG=746A37FE329B48A9922870D343DAB2B1&CID=0CAFD34A11B26EA30FA3D8B2104F6F1E&rd=1&h=FZfALwVHxE_S9H2WNKhtElNoV47kgQDk68vgEU4&v=1&r=https://www.nvidia.com/en-us/about-nvidia/ai-computing/&p=DevEx.LB.1,5549.1. [Accessed: 23-May-2018].