Color image segmentation based on region growing algorithm

Other Title(s)

تقسيم الصور الملونة بالاعتماد على خوارزمية التدرجات اللونية

Dissertant

al-Frahid, Muhammad Ibrahim

Thesis advisor

al-Sultani, Yas A.

Comitee Members

Kasasibah, Basil
al-Hamami, Ala H.
al-Bayatti, Hilal

University

Amman Arab University

Faculty

Collage of Computer Sciences and Informatics

Department

Department of Computer Science

University Country

Jordan

Degree

Master

Degree Date

2007

English Abstract

-Image segmentation is to divide the image into disjoint homogenous regions or classes, where all the pixels in the same class must have some common characteristics.

The proposed first technique is proposed to segment regions in an image, which is based on determining seed regions in an image.

The proposed first technique was applied on different sources of images with different complexities.

The segmentation was applied on each band of the RGB images, and also tested on the HSI images, the results showed that, we can obtain good segmentation by processing only the intensity band (I) of the HSI model, which gives similar results as that obtained in RGB model.

The results also showed the relationship between the number of dilation to close the regions and the number of regions that can be segmented.

Good results were obtained with the number of dilation equal 2, and the results showed that the best value to expand pixel neighbor is from 1 to 4, to keep the image from degration.

The proposed second technique (Part a) is proposed using differencing-averaging method to determine the image edges, the results showed that the first level of transformation is enough to give good results.

The edges of the image were enhanced, by using thresholding technique, followed by dilation operation to determine the closed segment regions.

The segments were colored with any color to discriminate these regions from the other regions of the image.

The medical eyes images are segmented by using two proposed techniques; the proposed second technique (Part b) and the proposed second technique (Part c).

The first is applied on iris image to determine edges of the iris, pupil and extract any different color regions, the technique was applied on two different sources of images and showed good results.

This technique is useful when we want to detect the white water inside the eyes.

The second consists of two stages, the first stage used the proposed second technique (part a) that used differencing-averaging method to determine the highlight region, and the second stage was based on drawing a circle to determine the object and it showed good results for the two different sources of medical images.

The background must be selected to obtain good results, by using background of high level of darkness.

The proposed third technique is proposed to segment color regions in an image that is based on determining edges and seed regions by using two-dimension wavelet transform.

By this technique, the image is enhanced by using Gaussian filter before it is transformed to first scale, and the noise that is set in the edges should be removed by using morphological algorithms.

The proposed third technique was applied on different types of images, the results showed the enhancement step is necessary to reduce the number of the seed regions, number of colored regions and elapsed time.

The results also showed that the best results were obtained when the value of the standard deviation for Gaussian filter was from 1 to 3.5, after that the image was degraded.

In the proposed third technique many of wavelet families were tested and the results showed that the best wavelet family was the first scale of the symlet to extract edges.

Main Subjects

Electronic engineering

No. of Pages

121

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-529532