A Survey of Shape Feature Extraction Techniques
A Survey of Shape Feature Extraction Techniques 57
4. Polygonal approximation
Polygonal approximation can be set to ignore the minor variations along the edge, and instead capture the overall shape information. This is useful because it reduces the effects of discrete pixelization of the contour. In general, there are two ways to realize it: one is merging methods and the other is splitting ones [22].
4.1 Merging methods
Merging methods add successive pixels to a line segment if each new pixel that is added doesn't cause the segment to deviate too much from a straight line.
4.1.1 Distance threshold method
Choose a point of the contour as a starting point. For each new point that we add, let a line go from the starting point to this new point. Then, compute the squared error for every point along the segment/line. If the error exceeds some threshold, we keep the line from the starting point to the previous point and start a new line at the current point.
In practice, the most of practical error measures in use are based on distance between vertices of the input curve and the approximation linear segments. The distance dk(i, j) from curve vertex Pk = (xk, yk) to the corresponding approximation linear segments (Pi, Pj) is

defined as follows (cf. Figure 12):

Fig. 12. Illustration of the distance from a point on the boundary to a linear segment
4.1.2 Tunneling method
If we have thick boundaries rather than single-pixel thick ones, we can still use a similar approach called tunneling. Imagine that we’re trying to lay straight rods along a curved tunnel, and that we want to use as few as possible. We can start at any point and lay as long a straight rod as possible. Eventually, the curvature of the “tunnel” won't let us go any further, so we lay another rod and another until we reach the end.
Both the distance threshold and tunneling methods can do polygonal approximation efficiently. However, the great disadvantage is that the position of the starting point will affect greatly the approximate polygon.
4.1.3 Polygon evolution
The basic idea of polygons evolution in [23] is very simple: in every evolution step, a pair of consecutive line segments (the line segment is the line between two consecutive vertices) s1,s2 is substituted with a single line segment joining their farther endpoints of s1 and s2.
ECG Feature Extraction Techniques - A Survey_生物学_自然科学_专业资料。ECG特征...intimately similar to the shape of the ECG signal achieved better detection...
A Survey of Shape Featur... 48页 20财富值 Submitted to Medicon 200......Feature Extraction Techniques CMU at TRECVID 2004Feature Extraction Techniques CMU...
Modified Fourier descriptor for shape feature extraction_专业资料。JCn.ot i(021:8—9 .etSuhUnv21)94845.DoI010/17102005:1.07s1 JCn...
Feature Extraction 6页 免费 feature extraction 暂无评价 12页 1财富值 3 Feature Extraction I 73页 1财富值 A Survey of Shape Featur... 48页 20财富值 ...
Iris Feature Extractio... 8人阅读 5页 2.00 Applying sub-band ener... 2人阅读 7页 2.00 Eigenvalue extraction ... 3人阅读 6页 2.00 ...
A Survey of Shape Feat... 48页 7下载券 A Survey...Vessel Extraction Techniques and Algorithms : A ...feature descriptions while snakes are used for ...
International Conference on Intelligent Information Hiding and Multimedia Signal Processing Shape Feature Extraction Using Fourier Descriptors with Brightness in Conte...
A Survey of Shape Featur... 48页 20财富值 ARS-Journal-SurveyPatter... ...Geometric Methods for Feature Extraction and Dimensional Reduction: A Guided Tou...
A Survey of Shape Featur... 48页 20财富值 An Exploration of the Pr.....A Survey of Answer Extraction Techniques in Factoid Question Answering Mengqiu...
A Survey of Shape Featur... 48页 20财富值 A Survey of Research on .....In industry, CAD techniques are extensively used to design products, and ...
热门文档