On chip implementation of a pixel-parallel approach for retinal vessel tree extraction

Abstract

Abstract-Retinal vessel tree extraction from angiography images plays an important role not only in the medical domain, but also in biometric identification applications. From the image processing point of view, many algorithms and strategies have been developed to deal with this topic. Although reliable results have been obtained, the main disadvantage in most of these proposals is still the high computation effort required.In this paper, a methodology to extract the retinal vessel tree has been developed, specially defined in terms of fine grain SIMD processing with the purpose of improving the computation time. The proposal has been implemented on a cellular processor array VLSI chip. The execution times for the main modules of the proposed algorithm have been included to show its capability.

Performance analysis of massively parallel embedded hardware architectures for retinal image processing

“…However, it affects to the performance and this is not generalizable and highlights the problems of these devices when their size is not easily scalable. More details of the implementation can be found in Alonso-Montes et al [30].…”

A cellular processor array simulation and hardware prototyping tool

“…The highly parallel nature of CPAs has lead to their increased use in a variety of fields, where many simple computations must be performed, and local connectivity would usually be a memory access bottleneck, such as vision chips [2], high-speed image processing [1] [5], cellular neural networks [3] and computational neuroscience models [6]. These applications only require simple numerical operations and neighbour communication, but use operations that are applied homogeneously to large data-sets, at high speeds.…”

Arteriolar-to-venular diameter ratio estimation: A pixel-parallel approach

“…Nevertheless, some of the proposed CNNbased operations utilise non linear templates preventing their implementation in the current generation of cellular processor VLSI chips. The proposal addressed in [17] has been implemented in terms of local convolutions and morphological operations together with arithmetic and logic operations, in order to be implemented in a fine grain single instruction multiple data (SIMD) parallel processors, particularly for the SCAMP-3 vision system [18]. This methodology finds the exterior of the vessels using an active contour technique, the so-called Pixel Level Snakes (PLS) [19].…”

“…In this paper, the retinal vessel tree extraction algorithm addressed in [17] is integrated into a medical application in order to compute the AVR ratio. A specialist classify the vessels (vein or artery) and the AVR ratio is automatically computed from the extracted retinal vessel tree.…”