Eng , Sheh Ling
(2016)
New Contrast Enhancement Technique For Non-Uniform Illumination Digital Colour Medical Images.
Masters thesis, Universiti Sains Malaysia.
Abstract
This dissertation presents a new non-linear contrast enhancement algorithm for non-uniform illumination and low contrast digital colour medical images. In this research study, medical microscopic cervical cell and human epithelial type 2 (HEp-2) cell images were employed as case studies. Commonly, the captured cell images from the video camera or digital camera have uneven illumination and poor contrast due to inadequate lighting, the quality of the image acquisition devices and/or environmental conditions. The problem of non-homogenous illumination is not considered by most of the developed contrast enhancement approaches when performing operations to improve the cell image quality. From the previous studies, although two non-linear dark and bright contrast enhancement methods were proposed, but each method was utilized to enhance the entire cell image. As a result, each resultant image contained only one enhanced region while degraded the contrast of another region extremely. Firstly, this proposed algorithm tackles the non-uniform illumination issue by implementing two modified Gaussian fuzzy membership functions to predetermined underexposed and overexposed regions. After obtaining more even illumination cell images, this proposed algorithm addresses the low contrast problem by proposing new non-linear dark region bright region contrast enhancement techniques to enhance dark and bright regions individually. Lastly, the enhanced pixels of each region are combined to form an enhanced image. According to the qualitative and quantitative analysis, the experimental results in greyscale and colour format showed that the proposed algorithm tends to provide clearer and informational enhanced images, least noise amplification, better differentiation between the cell and the background, better contrast and illumination, and capable to preserve the image naturalness as compared with other methods.
Actions (login required)
 |
View Item |
![[img]](http://eprints.usm.my/style/images/fileicons/application_pdf.png)
