• PDF / 1,727,558 Bytes
• 7 Pages / 439.37 x 666.142 pts Page_size
• 18 Downloads / 159 Views

DOWNLOAD

REPORT

---

Abstract High speed image processing is becoming increasingly important in medical imaging. Using the state-of-the-art ZYNQ-7000 system on chip (SoC) has made it possible to design powerful vision systems running software on an ARM processor and accelerating it from hardware resources on a single chip. In this paper, we take the advantage of accelerating an embedded system design on a single SoC, which offers the required features for real-time processing of skin cancer images. Different edge detection approaches such as Sobel, Kirsch, Canny and LoG have been implemented on ZYNQ-7000 for border detection of skin lesions, which can be used in early diagnosis of melanoma. The results show that the extended 5 9 5 canny edge detection implemented on the proposed embedded platform has better performance in compare with other reported methods. The performance evaluation of this approach has shown good processing time of 60 fps for real time applications. Keywords Border detection

 Edge detection  ZYNQ-7000  Medical imaging

Introduction To date, skin cancers have been one of the most common form of cancers particularly in New Zealand [1, 2]. Skin cancers are divided into two main categories: melanoma and non-melanoma. Early diagnosis of malignant melanoma can P. Sabouri (&)  H. GholamHosseini  J. Collins Department of Electrical and Electronics Engineering, Auckland University of Technology, Private Bag 92006, Auckland 1142, New Zealand e-mail: [email protected] H. GholamHosseini e-mail: [email protected] J. Collins e-mail: [email protected]

Y.-M. Huang et al. (eds.), Advanced Technologies, Embedded and Multimedia for Human-centric Computing, Lecture Notes in Electrical Engineering 260, DOI: 10.1007/978-94-007-7262-5_53,  Springer Science+Business Media Dordrecht 2014

465

466

P. Sabouri et al.

significantly decrease the morbidity, death and cost of the treatments [3]. Dermoscopy is a non-invasive method for diagnosis of melanoma and pigmented skin lesions. Although this device illustrates features of pigmented lesions, it is a challenging task for dermatologist to diagnose melanoma from other skin lesions [4]. Image processing techniques can be applied to skin images for better diagnosis of melanoma. For example, image features such as Asymmetry, Border irregularity, Color variation and regions with Diameter greater than 6 mm (ABCD rule) can be extracted using high performance image processing techniques [5]. In recent years, real-time vision systems have been used in a wide range of applications such as sophisticated medical imaging. Most computer-based vision applications have been developed based on a Graphics Processing Unit (GPU). However, recent developments in the field of powerful, low cost and energy-efficient embedded systems have led to the implementation of image/video applications into Digital Signal Processors (DSPs) and Field Programmable Gate Arrays (FPGAs). Nonetheless, each technique has its advantages and tradeoffs based on the nature of the algorithm, performance requirement, power con