Speed Improvement of B-Snake Algorithm Using Dynamic Programming Optimization

• Published in: IEEE transactions on image processing Vol. 20; no. 10; pp. 2848 - 2855
• Main Authors: Charfi, M., Zrida, J.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.10.2011; Institute of Electrical and Electronics Engineers; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Applied sciences; B-snake; Biological and medical sciences; Complexity; Computerized, statistical medical data processing and models in biomedicine; contours detection; Convergence; cubic B-splines; Dynamic programming; dynamic programming (DP); Exact sciences and technology; Femur - diagnostic imaging; Humans; Image processing; Image Processing, Computer-Assisted - methods; Information, signal and communications theory; Lung - anatomy & histology; Magnetic Resonance Imaging - methods; medical images; Medical management aid. Diagnosis aid; Medical sciences; Minimization; Noise; Optimization; Pattern recognition; Pixel; Searching; Shape; Signal processing; Spline; Telecommunications and information theory; Tomography, X-Ray Computed - methods; Upper bound; B spline; medical images; dynamic programming (DP); Image processing; cubic B-splines; Magnetic resonance; Shape detection; Algorithm; Computational complexity; Cubic spline; Tomography; B-snake; Medical imagery; Dynamic programming; contours detection; Edge detection; Active contour
• ISSN: 1057-7149; 1941-0042; 1941-0042
• DOI: 10.1109/TIP.2011.2134857

This paper presents a novel approach to contour approximation carried out by means of the B-snake algorithm and the dynamic programming (DP) optimization technique. Using the proposed strategy for contour point search procedure, computing complexity is reduced to O(N × M 2 ), whereas the standard DP method has an O(N × M 4 ) complexity, with N being the number of contour sample points and M being the number of candidates in the search space. The storage requirement was also decreased from N × M 3 to N × M memory elements. Some experiments on noise corrupted synthetic image, magnetic resonance, and computer tomography medical images have shown that the proposed approach results are equivalent to those obtained by the standard DP algorithm.