Segmentation in digital image processing

      Introduction to Image Segmentation                                            Technique

Image Segmentation

 Image segmentation is the process of partitioning a digital image into multiple segments. The main aim of segmentation is simplification. Representing an image into meaningful and easily analyzable way.  The goal of image segmentation is to divide an image into multiple segments based on similar features or attributes. Image segmentation is used to locate object and boundaries in images.

Segmentation algorithm based on of two basic properties of gray-scale values:

           1. Discontinuity: This is the approach in which an image is segmented into regions based on discontinuity. It partition an image based on abrupt changes in gray scale levels. The edge detection based segmentation belong in this category in which edges formed due to intensity discontinuity are detected and linked to form boundaries of regions.

           2. Similarity: The techniques that based on this approach are: Thresholding techniques, region growing techniques and region splitting and merging. These all divide the image into regions having similar set of pixels.  

                         

In other words, also we can say that image segmentation can be approached from three perspectives: Region approach, Edge approach and Data clustering. The region approach falls under similarity detection and edge detection and boundary detection falls under discontinuity detection. Clustering techniques are also under similarity detection

We can partition ta image into various parts also called segments. After dividing an image into segments, instead of processing the entire image you can process only the important segments of the image.

The Different Types of Image Segmentation

      1.     Thresholding Method

2.     Region-based Segmentation

3.     Edge Detection Segmentation

4.     Image Segmentation based on Clustering

1.    Thresholding Method

In this method an image is partitioned into two or more sub-images upon comparing with the predefined threshold value. These methods divide the image pixel accordingly to their intensity level. These methods are used over images having lighter objects than background. Suppose that an image, f(x, y), is composed of light objects on a dark background, and the following figure is the histogram of the image.

Then, the objects can be extracted by comparing pixel values with a threshold T. One way to extract the objects from the background is to select a threshold T that separates object from background. Any point (x, y) for which f(x, y) > T is called an object point; otherwise the point is called a background point.g (x, y) =    ^{〘  1        if         f (x, y) > T  f}

         ^{0   f (x, y) < = T}

^{When T is a constant applicable over an entire image, then the above process is called as Global thresholding.}

^{When the value of T changes over an image then that      process is referred as Variable thresholding. Sometimes it is also termed as local or regional thresholding. Where, the value of T at any point (x ,y) in an image depends  on properties of a neighborhood of (x, y) .If T depends on the spatial coordinates (x ,y) themselves, then  variable thresholding is often referred to as dynamic or  adaptive thresholding.}

  

2.    Region-based Segmentation

Region-based segmentation divide an image into regions. This method work on the principle of homogeneity. This technique partition an image into several regions based on criteria like color, intensity, object.

Types of region based:

1.      Region Growing - Region growing is a procedure that groups pixel or sub-regions into larger regions. It starts with a set of “seed” points and from these grows regions by appending to each seed points those neighboring pixels that have similar properties (such as gray level, texture, color, shape).In noisy image edges are extremely difficult to detect. For noisy image region growing is better. Homogeneity of regions is most important in segmentation. There are several criteria for homogeneity: gray level, color, texture, shape, model etc.        

2.      Region Splitting- An alternative is to start with the whole image as a single region and subdivide the regions that do not fulfill a condition of homogeneity.

3.      Region Merging-   Region merging is the opposite of region splitting. Start with small regions like 2*2 or 4*4 regions and marge the regions that have similar characteristics. Splitting and merging approaches are used iteratively.

4.      Split and Merge with Quad Trees

Initialize segmentation is arbitrarily or using prior knowledge.

For each region R, if it is not homogeneous, split it. This builds a quad tree.

If R1 and R2 are neighbors, and they can be merged into a homogeneous region do it. This might destroy the quad tree structure, because you might get a1/4 region connected to a 1/6 region.

If any regions are too small merge them with the best neighbor.

                                                                                                                

  

Quad tree

1.      Split into 4 disjoint quadrants any region R_i  for which  P(R_i) = FALSE

2.      Merge any adjacent region R_j and R_k for which

P (R_i <= R_k ) = TRUE

Stop when no further merging or splitting is possible.

3.    Edge Detection Segmentation

Edge detection is an image processing technique for detecting the boundaries of objects. It works by finding discontinuities in brightness. In this method first of all the edges are detected and then are connected together to form the object boundaries to segment the required regions.

This methods transform original images into edge images.  It is a fundamental process detects and outlines of an object and boundaries among objects and the background in the image. Edge detection is the most familiar approach for detecting significant discontinuities in intensity values. A single intensity value does not provide good information about edges. Edge detection techniques locate the edges where either the first derivative of intensity is greater than a particular threshold or the second derivative has zero crossings.

There are two basic methods first is Gray histograms based method and second is Gradient based methods. For detecting edges there are several techniques are used like canny operator, sobel operator and Robert’s operator etc. Resulting images will be binary image. These are the structural techniques based on discontinuity detection.

Image Segmentation based on Clustering

Clustering is unsupervised learning method. It deals with finding a structure in a collection of unlabeled data. Clustering is the process of organizing objects into groups accordingly similar properties of object. Similar cluster are group together.

There are several methods and one of the most popular methods is K-Means clustering algorithm.

K-means algorithm is simple and easy to understand. For small set k-means are very suitable. This is an unsupervised algorithm. It is used to segment the interest area from the background. Based on the K-centroids it clusters or partitions the given data into K-clusters or parts. We assign the points to the clusters which are closest to them. K-means is a distance based algorithm.

Steps of K-Means algorithm:

1. Choose the number of clusters K.

2. Select at random K points, the centroids (not necessarily from your dataset).

3. Assign each data point to the closest centroid → that forms K clusters.

4. Compute and place the new centroid of each cluster.

5. Reassign each data point to the new closest centroid. If any reassignment took place, go to step (4), otherwise, the model is ready.

Applications

1.      Medical Images i.e. finding tumors veins, etc.

2.      Satellite/Aerial images i.e. finding targets  

3.      Object Detection i.e. face recognize, fingerprint recognize

4.      Surveillance Images

5.      Video surveillance

CONCLUSION

Image segmentation is one of the most important method. Segmentation is used to partition an image into several disjoint subsets called segments. Each segment corresponds to a meaningful part of the image. Image segmentation is a difficult task since often the scene objects are defined by image regions with non-homogenous texture and color characteristics. Although a variety of image segmentation methods have been developed, there is still no general method suitable for segmenting any type of images and thus research on image segmentation method has become the key issue. Therefore, further research is required to develop an effective segmentation technique to partition an image for locating and identifying objects or boundaries in an image in a clear and meaningful way to meet the demands of advanced applications.  These techniques can be used for object recognition and detection. In medical images these can be used to detect cancer and in satellite images these can be used to detect roads and bridges. Thus it is clear that various methods are suitable for various types of image applications. But from the study it is clear that no single method is sufficient for every image type and no all methods are suitable for a particular image type. Due to the need of image segmentation in many applications, it has a challenging future.