Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers Motivation Contour detection in a restricted search space is: More reliable Fast Institute for Computer Science, University of Münster Department of Nuclear Medicine, University Hospital Münster Computer Tomograph imagePositron Emission Tomograph image
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 2 Restricting the Search Space Bounding box
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 3 Restricting the Search Space Bounding box Dual snakes
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 4 Restricting the Search Space Bounding box Dual snakes Surface normals Our approach: Dual band
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 5 The procedure Step 1: Form the dual band from the initial contour Step 2: Interlink the dual band boundaries Step 3: Find the target contour within the dual band
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 6 Dynamic Programming Optimal solutions of sub-problems add up to global optimum, provided the sub-problems depend upon the steps before them only and there is sequence of sub-problems. Global optimum Fast Non-iterative
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 7 Step 1: Defining the Dual-Band Our Approach: Morphological dilation and erosion
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 8 Step 1: Defining the Dual-Band Our Approach: Morphological dilation and erosion
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 9 Step 2: Defining the links Use dynamic programming to find the best match between points on the dual contour Use the sum of lengths of the connecting lines as cost function
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 10 Step 2: Defining the links Use dynamic programming to find the best match between points on the dual contour Use the sum of lengths of the connecting lines as cost function Inner Contour Outer Contour
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 11 Step 3: Finding the Contour in the Dual-Band Use the connecting lines as rows of the image matrix Define a cost function to find the best contour ………… ImageImage Matrix ………… Cost Matrix
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 12 Experimental Results Tracking
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 13 Experimental Results Segmenting lungs on PET images
Reliable Dual-Band Based Contour Detection A Double Dynamic Programming Approach Mohammad Dawood, Xiaoyi Jiang, Klaus P Schäfers 14 Conclusion and Future Work We have developed a: fast non-iterative robust procedure for contour detection in restricted search space Work on 3D extension is already underway.