Conference Proceeding

Object segmentation based on split and merge algorithm

Dept. of Comput. Sci. & Eng., Mil. Inst. of Sci. & Technol., Dhaka
12/2008; DOI:10.1109/TENCON.2008.4766802 In proceeding of: TENCON 2008 - 2008 IEEE Region 10 Conference
Source: IEEE Xplore

ABSTRACT Image segmentation is a feverish issue as it is a challenging job and most digital imaging applications require it as a preprocessing step. Among various algorithms, although split and merge (SM) algorithm is highly lucrative because of its simplicity and effectiveness in segmenting homogeneous regions, however, it is unable to segment all types of objects in an image using a general framework due to not most natural objects being homogeneous. Addressing this issue, a new algorithm namely object segmentation based on split and merge algorithm (OSSM) is proposed in this paper considering image feature stability, inter- and intra-object variability, and human visual perception. The qualitative analysis has been conducted and the segmentation results are compared with the basic SM algorithm and a shape-based fuzzy clustering algorithm namely object based image segmentation using fuzzy clustering (OSF). The OSSM algorithm outperforms both the SM and the OSF algorithms and hence increases the application area of segmentation algorithms.

0 0
 · 
0 Bookmarks
 · 
51 Views
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: A hybrid multidimensional image segmentation al- gorithm is proposed, which combines edge and region-based techniques through the morphological algorithm of watersheds. An edge-preserving statistical noise reduction approach is used as a preprocessing stage in order to compute an accurate estimate of the image gradient. Then, an initial partitioning of the image into primitive regions is produced by applying the watershed trans- form on the image gradient magnitude. This initial segmentation is the input to a computationally efficient hierarchical (bottom- up) region merging process that produces the final segmentation. The latter process uses the region adjacency graph (RAG) repre- sentation of the image regions. At each step, the most similar pair of regions is determined (minimum cost RAG edge), the regions are merged and the RAG is updated. Traditionally, the above is implemented by storing all RAG edges in a priority queue. We propose a significantly faster algorithm, which additionally maintains the so-called nearest neighbor graph, due to which the priority queue size and processing time are drastically reduced. The final segmentation provides, due to the RAG, one-pixel wide, closed, and accurately localized contours/surfaces. Experimental results obtained with two-dimensional/three-dimensional (2-D/3- D) magnetic resonance images are presented.
    IEEE Transactions on Image Processing. 01/1998; 7:1684-1699.
  • [show abstract] [hide abstract]
    ABSTRACT: A new clustering algorithm is developed for efficient classification of data in μ when there exists no a priori information about the number of clusters. The algorithm is based on a split-and-merge technique. The type-I splitting is guided by density of data over strips at different directions around the centroid of the data. The type-II splitting is the usual K-means clustering algorithm (K = 2) and rechecked with the help of a merging technique. A theorem on the convergence of this algorithm is proved.
    Pattern Recognition Letters. 01/1992; 13:399-409.
  • Source
    [show abstract] [hide abstract]
    ABSTRACT: The image segmentation performance of clustering algorithms is highly dependent on the features used and the type of objects contained in the image, which limits the generalization ability of such algorithms. As a consequence, a fuzzy image segmentation using suppressed fuzzy c-means clustering (FSSC) algorithm was proposed that merged the initially segmented regions produced by a fuzzy clustering algorithm, using two different feature sets each comprising two features from pixel location, pixel intensity and a combination of both, which considered objects with similar surface variations (SSV), the arbitrariness of fuzzy c-means (FCM) algorithm using pixel location and the connectedness property of objects. The feature set selection for the initial segmentation in the merging technique was however, inaccurate because it did not consider all possible feature set combinations and also manually defined the threshold used to identify objects having SSV. To overcome these limitations, a new automatic feature set selection for merging image segmentation results using fuzzy clustering (AFMSF) algorithm is proposed, which considers the best feature set selection and also calculates the threshold based upon human visual perception. Both qualitative and quantitative analysis prove the superiority of AFMSF algorithm compared with other clustering techniques including FSSC, FCM, possibilistic c-means (PCM) and SFCM, for different image types.

Full-text

View
0 Downloads
Available from