Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
1K views

On Color Image Segmentation

This document discusses using a pillar k-means clustering algorithm for color image segmentation. It begins by providing background on image segmentation and some common algorithms. It then describes the traditional k-means clustering algorithm process and some drawbacks. The document introduces the pillar k-means clustering algorithm as an optimization of k-means that positions centroids farther apart to improve precision, reduce computation time, and enhance segmentation quality. The pillar algorithm designates initial centroids based on maximum distance between data points to provide optimized starting positions for k-means clustering.

Uploaded by

rajeshbabu_nn641
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
1K views

On Color Image Segmentation

This document discusses using a pillar k-means clustering algorithm for color image segmentation. It begins by providing background on image segmentation and some common algorithms. It then describes the traditional k-means clustering algorithm process and some drawbacks. The document introduces the pillar k-means clustering algorithm as an optimization of k-means that positions centroids farther apart to improve precision, reduce computation time, and enhance segmentation quality. The pillar algorithm designates initial centroids based on maximum distance between data points to provide optimized starting positions for k-means clustering.

Uploaded by

rajeshbabu_nn641
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 17

Color image segmentation using pillar k-means clustering algorithm

BY A R S BALAJI, R.NO: 09H91D6302 M.TECH(DIP).

Image segmentation
Image segmentation plays a vital role in image processing Image segmentation sub-divides the image into regions or objects, in order to extract interesting parts of an image such as color, texture, shape and structure. Attributes
Image SEGMENTATION of an image

Color Image segmentation


Image segmentation can be extended to colored images, which has been extensively applied in face recognition, location of satellite images, medical imaging. There are several algorithms are developed for segmenting gray scale images, But segmenting the color images has received very less attentions to the scientific community. It is difficult to segment an image with both color & texture.

Segmentation Algorithms

K-Means clustering algorithm


Properties :
K-Means clustering algorithm is unsupervised, non-deterministic and iterative. There are always k clusters. Each cluster is having at least one item. The clusters are non-hierarchical and they dont overlap.

The K-Means clustering algorithm process

the data set is partitioned into K-clusters & data points are randomly assigned to k-clusters. For each data point, calculate the shortest distance from the data point to each cluster, which can be calculated by using Euclidean distance.

If the data point is close to its cluster leave it where it is. If the data point is not closest to its own cluster, move it into the closest cluster. Repeat this step until it forms data matrix i.e., no data point moving from one cluster to the another.

The K-Means clustering algorithm process

Drawbacks of K-Means
In the K-means Algorithm images were segmented
neatly. But It is a TIME TAKING process because of the iterations procedure that it performs. K-means algorithm is difficult to reach global optimum, but only to one of local minima which it will lead to incorrect results. In order to optimize K-means clustering for image segmentation, we propose Pillar algorithm.

Pillar K-Means clustering


The Pillar algorithm is very robust and superior algorithm It provides optimization for K-means by positioning all centroids far separately among them in the data distribution. The segmentation process by our approach includes a new mechanism for clustering such as
Improve precision Reduce computation time and Enhance the quality of image segmentation.

Pillar K-Means clustering


Calculate the mean Normalization Calculate the distance between data points & mean and consider maximum distance between them

Block diagram representation of k-means clustering:

An adaptive noise removal filtering using the Wiener filter is applied for noise removal of images. Image segmentation system pre-proceeds the image by transforming the color space from RGB to HSL and CIELAB color systems. Here we utilize both of them as hybrid color systems for image segmentation. Because of different ranges of data points in HSL and CIELAB color spaces, we need to normalize the datasets by using Softmax algorithm.

Procedure for pillar (optimization of) k-means clustering:


Designate the initial centroids positions by calculating the farthest distance between data points and previous centroids. Selects data points which have the maximum distance as new initial centroids.

Figure illustrating the location of set of pillars (white points)

After getting optimized initial centroids, we can obtain final centroids by applying clustering using k-means.

Thus this mechanism is able to improve segmentation results and make faster computation for the image segmentation when compared to k-means clustering algorithm.

Thank you

You might also like