An Improved Denoised 3D Edge Extraction Operator within Biomedical Images - Multimedia and Signal Processing - page 309

An Improved Denoised 3D Edge Extraction Operator
within Biomedical Images

Yu Ma 1,2 , Yanning Zhang 1 , Yougang Wang 2 , Huilin Liang 2 , and Shuang Xu 2

1 Institute of Computer Science, Northwestern Polytechnical University,
Xi'an, 710129, China
2 Ning Xia University, Yin chuan, 750021, China
[e-mail]

Abstract. The previous 3D edge surface detector based on the Laplace and
gradient operator can extract high accuracy edge surfaces with high efficiency in
contrast with traditional isotropic surface extraction operator. However, the
second derivat1ive in the 3D detector shows natural sensitivity to noise, which
generates the noise polluted 3D edge surfaces and noisy pieces. A novel
denoising 3D edge detector is proposed; the noisy image is filtered by the 3D
Gauss filter firstly, then edge surfaces are detected and extracted utilizing the
traditional 3D edge surface detector. Furthermore, the extracted 3D noisy edge
surface pieces are degraded by the tracking technique. Finally, the denoising 3D
edge surfaces are converted to polygon pieces, then visualized the surface with
combined image and graphic methods. Experimental results show that the
proposed scheme suppresses noise and preserves edge surfaces than the
traditional 3D edge surface detector.

Keywords: 3D edge surface detector, 3D Gaussian filter, denoising, tracking.

1 Introduction

Computed tomography (CT) and magnetic resonance imaging (MRI) produce the
image sequence of two-dimensional (2D) cross-sectional slices which contain the
three-dimensional (3D) information on the biomedical and industrial object [1].
However, the accompanying noise during image acquisition degrades the human
interpretation or computer-aided analysis of the images. In a 3D reconstruction of
biomedical images, it is a key step to extract the edge surfaces of the interesting organ
or tissue from the 2D slices. However, the noises would lead to wrong or misleading
edge surface structures in the images of the object. Therefore, denoising should be
performed to improve the edge surface quality for more accurate diagnosis and
applications. In this paper, we exploit a novel edge surface extraction detector based
denoising in 3D volume data.
The Canny operator [2] is a classical traditional operator in 2D edge detection, which
computes local maximum based gradient. The 3D edge detection operator is the
extension of the 2D Canny Detector. In recent years many reports on the detection of the
edge surface within 3D volume images have appeared in the literature [3], [4], [5], [6].