Real-time thinning algorithms for 2D and 3D images using GPU processors
- PDF / 2,091,317 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 109 Downloads / 188 Views
ORIGINAL RESEARCH PAPER
Real‑time thinning algorithms for 2D and 3D images using GPU processors Martin G. Wagner1 Received: 31 October 2018 / Accepted: 22 May 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019
Abstract The skeletonization of binary images is a common task in many image processing and machine learning applications. Some of these applications require very fast image processing. We propose novel techniques for efficient 2D and 3D thinning of binary images using GPU processors. The algorithms use bit-encoded binary images to process multiple points simultaneously in each thread. The simpleness of a point is determined based on Boolean algebra using only bitwise logical operators. This avoids computationally expensive decoding and encoding steps and allows for additional parallelization. The 2D algorithm is evaluated using a data set of handwritten characters images. It required an average computation time of 3.53 ns for 32 × 32 pixels and 0.25 ms for 1024 × 1024 pixels. This is 52–18,380 times faster than a multi-threaded border-parallel algorithm. The 3D algorithm was evaluated based on clinical images of the human vasculature and required computation times of 0.27 ms for 128 × 128 × 128 voxels and 20.32 ms for 512 × 512 × 512 voxels, which is 32–46 times faster than the compared border-sequential algorithm using the same GPU processor. The proposed techniques enable efficient real-time 2D and 3D skeletonization of binary images, which could improve the performance of many existing machine learning applications. Keywords Centerline · GPU Programming · Medial axis · Skeletonization · Thinning
1 Introduction The skeletonization of binary images is a common task in image processing applications. The resulting skeleton, often also referred to as medial axis or medial lines, reduces the complexity of an object while retaining its topology. Arguably, one of the most common application is optical character recognition (OCR), where handwritten or printed letters are classified based on their skeletons [1, 2]. In [3–5] skeletons are used for automatic chromosome analysis (karyotyping), to identify abnormalities in the morphology. Also for computer vision tasks such as object recognition [6, 7] and object tracking [8] skeletonization is an important tool. Furthermore, in [9] it is used to reconstruct the 3D path of interventional devices from two orthogonal projection images. Other applications include extraction of features for fingerprint analysis [10] and protein structure analysis [11]. Many of the tasks and applications that use skeletonization * Martin G. Wagner [email protected] 1
Department of Medical Physics, University of Wisconsin, 1111 Highland Avenue, Madison, WI 53705, USA
require real-time image processing. The focus of this work is to propose a fast and accurate thinning algorithm that is suitable for these tasks and discuss its efficient implementation on modern GPUs. Because of the wide variety of applications that rely on skeletonization, a large number of publication
Data Loading...
