China
Africa
Explore chapters and articles related to this topic
Filtering Algorithm
Published in Ahmad Fikri Bin Abdullah, A Methodology for Processing Raw Lidar Data to Support Urban Flood Modelling Framework, 2020
The third group of algorithms includes those that progressively increase the density of the points for the DTM in order to approximate the bare earth (e.g., Elmqvist, 2002). The ground surface is determined by employing an active shape model. A deformable model is used to fit the bare earth by progressively minimising the energy associated with the active shape model. When applied to the LiDAR data, the active shape model behaves like a membrane, floating up from underneath the data points. The manner in which the membrane sticks to the data points is determined by an energy function. For the membrane to stick to the ground points, it has to be chosen in such a way, that the energy function is minimized. Hu (2003), Wack and Wimmer (2002), and Pfeifer and Stadler (2001) used a hierarchical approach, which is similar to the pyramid method used by Adelson et al., (1984). In this approach, a coarse DTM is generated at the top level first and then refined hierarchically.
Segmentation and Classification
Published in Elizabeth Berry, A Practical Approach to Medical Image Processing, 2007
Figure 2.11 shows a snake being used on an x-ray image of the hand. The segmentation task is to find the edges of the largest bone in the image. Figure 2.11a shows the initial ellipse-shaped contour, and Figure 2.11b is the result after the snake program has run. The contour has successfully moved to the edges of the bone, but, because there are no shape constraints, at the ends of the bone, it has been attracted to an edge that is part of a different bone. This characteristic of snakes means that the initial contour (the ellipse on the left in this case) needs to be placed very close to the correct edge to start with. The deformable, or active shape, model overcomes this drawback by including knowledge of the shape that is expected. In the example in Figure 2.12, the initial contour (Figure 2.12a) is clearly bone shaped. In the resulting image after the program has been run (Figure 2.12b), the shape has fitted itself neatly around the edges of the bone.
Multimedia-Based Affective Human–Computer Interaction
Published in Ling Guan, Yifeng He, Sun-Yuan Kung, Multimedia Image and Video Processing, 2012
Yisu Zhao, Marius D. Cordea, Emil M. Petriu, Thomas E. Whalen
Feature-based methods localize the facial features of an analytic face model in the input image or track them in the image sequence. Model-based methods fit a holistic face model to the face in the input image or track it in the image sequence. Deformable templates in general and active models (Active Shape Model (ASM), Active Appearance Model (AAM)) [27] in particular represent good examples of the holistic approach to extract facial features.
A survey on computer vision techniques for detecting facial features towards the early diagnosis of mild cognitive impairment in the elderly
Published in Systems Science & Control Engineering, 2019
Zixiang Fei, Erfu Yang, David Day-Uei Li, Stephen Butler, Winifred Ijomah, Huiyu Zhou
A promising approach to face alignment, involving aspects of both methods, was proposed by Cootes, Taylor, Cooper, and Graham (1995). The Active Shape Model (ASM) algorithm contains a global shape model and many local feature models. The major steps for the ASM algorithm to find the best facial landmarks can essentially be summarized as follows (Cootes et al., 1995; Huang et al., 2010): The shape parameter b is initialized to zero.The shape model points are generated by , where is the mean shape model, P is the eigenvectors corresponding to the largest eigenvalues and b is the shape parameter.The best landmark z is found by the feature model.The shape parameter is obtained by .The shape parameter is restricted within .