A Level Set Approach for Segmentation of Intensity Inhomogeneous Image Based on Region Decomposition
- PDF / 2,359,273 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 92 Downloads / 221 Views
ORIGINAL RESEARCH
A Level Set Approach for Segmentation of Intensity Inhomogeneous Image Based on Region Decomposition Deepa Chakravarty1 · Debasish Pradhan1 Received: 13 January 2020 / Accepted: 30 July 2020 © Springer Nature Singapore Pte Ltd 2020
Abstract Medical images like magnetic resonance angiogram (MRA) images have inhomogeneous intensities. When such images are segmented using active contour models, they do not produce desired results. Hence, segmenting an image with inhomogeneous intensities is a challenge. In this paper, the problem of segmenting medical images with intensity inhomogeneity has been addressed. Here, we propose a method which uses the concept of region decomposition along with combination of region-based segmentation and edge-detector function. In this process, the input image is divided into four quadrants and checked for segmentation. If the segmentation of any sub-image is not as desired, it is further decomposed into four more sub-regions. The process continues until each sub-image is properly segmented. In the end, all the segmented parts are merged to obtain the desired segmented image. The proposed method has been implemented and desired results are obtained by the same and compared with existing results. Keywords Segmentation · Active contours · Level sets · Intensity inhomogeneity · Region splitting and merging · Edge detector · Partial differential equation
Introduction Segmentation divides an image into foreground and background. The problem of image segmentation is of utmost importance nowadays. Image segmentation has found its significance in various fields of science and technology; few to be named are computer vision, medical imaging, object detection, traffic control, video surveillance. Image segmentation can be broadly classified into two categories, namely, (1) edge-based segmentation, and (2) region-based segmentation. Edge-based segmentation model [5, 6, 13, 14] makes use of the image gradient and it may also involve an edge-detector function to converge the active contour towards the edge of the desired object in the image. On the other hand, a region-based segmentation model [15, 24, 26] does not use the gradient information of the image but utilizes statistical information based on image intensity.
* Debasish Pradhan [email protected] 1
Department of Applied Mathematics, Defence Institute of Advanced Technology, Girinagar, Pune 411025, Maharashtra, India
One of the most efficient schemes used to detect smooth boundaries of an object in an image was proposed by Kass et al. [13]. In this model, the authors have used the information of image gradient to attract active contour to the edges of the object. Ever since, several models of image segmentation have been extensively studied. Later, Mumford et al. [18] proposed an active contour model based on a region-based approach. This model is popularly known as the Mumford–Shah (MS) model. Several models [8, 16, 20, 24] have been proposed based on the Mumford–Shah model to improve the detection of an object i
Data Loading...
