An Overview of image segmentation (I.)

An Overview of image segmentation (I.)

[Email protected]

Http://blog.csdn.net/zouxy09

The so-called image segmentation refers to the gray, color, texture and shape characteristics of the image into a number of non-overlapping areas, and make these features in the same region to show similarity, and in different regions show obvious differences. We first to the current main image segmentation method to do an overview, and then the individual methods to do a detailed understanding and learning.

1, Threshold-based segmentation method

The basic idea of the threshold method is to calculate one or more gray threshold values based on the gray feature of the image, and compare the gray value of each pixel in the image with the threshold value, and then divide the pixels into the appropriate category according to the comparison results. Therefore, one of the most critical steps in this class of methods is to solve the best gray threshold values according to a criterion function.

2, Edge-based segmentation method

The so-called Edge refers to the set of successive pixels on the boundary line of two different regions in the image, and is the reflection of the discontinuity of the local feature of the image, which reflects the mutation of the image characteristics such as grayscale, color and texture. Usually, edge-based segmentation refers to the edge detection based on gray value, which is based on the observation that the edge gray value will show the change of step or roof type.

There are obvious differences between the gray values of the pixel points on the edge of the step type, while the roof shape edge is at the turning point where the gray value rises or falls. Based on this feature, it is possible to use differential operators for edge detection, even with the extremum of the first derivative and the 0 points of the second derivative to determine the edge, which can be accomplished using the image and template convolution.

3, Region-based segmentation method

The method is divided into different regions according to the similarity criterion, including seed region growth method, regional division merging method and watershed method.

The seed region growth method starts with a group of seed pixels representing different growth regions, then merges the eligible pixels in the seed pixel neighborhood into the growth region represented by the seed pixel, and continues the merge process as a new seed pixel until a new pixel matching the criteria is found. The key of this method is to select the appropriate initial seed pixel and reasonable growth criterion.

The basic idea of the regional division merging Method (GONZALEZ,2002) is to divide the image into several disjoint regions first, then divide or merge the regions according to the relevant criteria, which is suitable for both gray image segmentation and texture image segmentation.

Watershed Method (meyer,1990) is a mathematical morphology segmentation method based on topological theory, whose basic idea is to think of images as topological landforms on geodetic topography, and the gray value of each pixel in the image represents the elevation of the point, each local minimum and its affected area is called the catchment basin. The boundary of the catchment basin is a watershed. The implementation of the algorithm can be simulated into the process of flooding, the lowest point of the image is submerged first, and then the water gradually drowns the entire valley. When the water level reaches a certain height will overflow, when the water overflow in the construction of the dam, repeat the process until the entire image of the point is flooded, then the establishment of a series of dams will be divided into separate basins of the watershed. The watershed algorithm has a good response to the weak edge, but the noise in the image will cause the watershed algorithm to be segmented.

4. The segmentation method based on graph theory

This method associates the problem of image segmentation with the minimum cut (min cut) problem of the graph. First, the image is mapped to a weighted g=,e>, and each node in the graph n∈v corresponds to each pixel in the image, each edge ∈e a pair of neighboring pixels, and the weight of the edge represents the nonnegative similarity between neighboring pixels in terms of grayscale, color, or texture. And a division of the image is a cut of the graph, each area of the partition C∈s corresponds to a sub-graph in the diagram. The optimal principle of segmentation is to maintain the similarity between the sub-graphs and the sub-graphs with the least similarity. The essence of the segmentation method based on graph theory is to remove the specific edges and divide the graph into several sub-graphs to realize the segmentation. At present, the method based on graph theory has graphcut,grabcut and random walk and so on.

5. Segmentation method based on energy functional function

The method mainly refers to the active contour model (active contour model) and the algorithm developed on it, the basic idea is to use continuous curve to express the target edge, and define an energy function to make its independent variable including The edge curve, Therefore, the segmentation process is transformed into the process of solving the minimum value of the energy functional, which can be solved by solving the Euler (Euler) function. Lagrange) equation to achieve the curve position where the energy reaches the hour is the contour of the target. The active contour model can be divided into two categories according to the form of curve expression in the model: the parametric active contour model (parametric active contour model) and the geometric active contour model (geometric active contour model).

The parametric active contour model is based on the Lagrange frame, which expresses the curve directly in the parametric form of the curve, and the most representative is the snake model proposed by Kasset A1 (1987). This kind of model has been successfully applied in the field of early biological image segmentation, but it has the disadvantage that the segmentation result is influenced by the setting of the initial contour and difficult to deal with the change of the topological structure of the curve, besides, its energy function depends only on the selection of the curve parameter, which is independent of the geometrical shape of the object.

The curve motion process of the geometric active contour model is based on the geometrical measurement parameters of the curve and not the expression parameters of the curve, so the change of topological structure can be handled well, and the problem that the parametric active contour model is difficult to solve is solved. The introduction of the level set method (osher,1988) has greatly promoted the development of the geometric active contour model, so the geometric active contour model can also be called the level set method.

An Overview of image segmentation (I.)