GPU-Based Fast Refinements for High-Quality Color Volume Rendering
Color volume datasets of the human body, such as Visible Human or Visible Korean, describe realistic anatomical structures. However, imperfect segmentation of these color volume datasets, which are typically generated manually or semi-automatically, produ
- PDF / 284,775 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 107 Downloads / 198 Views
Abstract Color volume datasets of the human body, such as Visible Human or Visible Korean, describe realistic anatomical structures. However, imperfect segmentation of these color volume datasets, which are typically generated manually or semi-automatically, produces poor-quality rendering results. We propose an interactive high-quality visualization method using GPU-based refinements to support the study of anatomical structures. To smoothly represent the boundaries of a region-of-interest (ROI), we apply Gaussian filtering to the opacity values of the color volume. Morphological grayscale erosion operations are performed to shrink the boundaries, which are expanded by the Gaussian filtering. We implement these operations on GPUs for the sake of fast refinements. As a result, our method delivered high-quality result images with smooth boundaries providing considerably faster refinements, sufficient for interactive renderings as the ROI changes, compared to CPU-based method. Keywords Color volume rendering
GPU-based refinement
B. Lee Research and Development Department, Zetta Imaging, Seoul, Republic of Korea e-mail: [email protected] K. Kwon B.-S. Shin (&) Department of Computer Science and Information Engineering, Inha University, Incheon, Republic of Korea e-mail: [email protected] K. Kwon e-mail: [email protected] © Springer Science+Business Media Singapore 2016 J.J.(Jong Hyuk) Park et al. (eds.), Advances in Parallel and Distributed Computing and Ubiquitous Services, Lecture Notes in Electrical Engineering 368, DOI 10.1007/978-981-10-0068-3_14
117
118
B. Lee et al.
1 Introduction Direct volume rendering (DVR) efficiently visualizes medical images such as computerized tomography or magnetic resonance images [1]. Since these images do not contain optical information, DVR normally classifies voxels using a transfer function to map the voxel values to colors. Color volume datasets such as Visible Human (VH) [2] and Visible Korean (VK) [3] data provide anatomical structure information. However, assigning a region-of-interest (ROI) to color volume datasets using an opacity transfer function (OTF) is difficult due to the weak correlation between color values and organs. Hence, it is difficult to render color volume data using a general DVR method [4]. Color values are generally transformed into a monochrome space using various color spaces when color images are segmented [5]. OTF can be applied to the CIE L*U*V color space transformed from the RGB color space [6]. However, these methods are effective only for specific organs having a similar color distribution. Segmentation data can be used as an additional volume data that contains ROI information. With this data, it is possible to render color volume datasets [7, 8]. However, segmentation is a time-consuming task because it is performed manually and these color volume datasets consist of thousands of slices. Furthermore, as illustrated in Fig. 1, the result may not be as precise due to complicated human anatomical structures. Since the segmentation data has a discrete
Data Loading...
