Adaptive Local Fuzzy Based Region Determination Image Enhancement Techniques For Non-Uniform Illumination And Low Contrast Images

• Main Author: Salih, Abdullah Amer Mohammed
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: T Technology; TK1-9971 Electrical engineering. Electronics. Nuclear engineering
• Online Access: http://eprints.usm.my/47840/1/Adaptive%20Local%20Fuzzy%20Based%20Region%20Determination%20Image%20Enhancement%20Techniques%20For%20Non-Uniform%20Illumination%20And%20Low%20Contrast%20Images.pdf; http://eprints.usm.my/47840/

Local contrast enhancement is an approach to improve the local visibility detail of an image by increasing the contrast in local regions. Recently, researchers have shown an interest in solving the issue of non-uniform illumination. However, most of these techniques divide the image into two parts only namely over-exposed and under-exposed regions and try to enhance the poor contrast in both regions using same approach. However, these methods are not robust and they are specifically designed to solve a specific problem at one time. This limitation has motivated this study to propose a new technique to solve the abovementioned problems. In the beginning, Adaptive Local Exposure Based Region Determination (ALEBRD) method is proposed to determine and divide the image into three regions namely under-exposed, over-exposed, and well-exposed regions. The results show that the proposed ALEBRD method produced better region determination performance than the other state-of-the-art methods. Based on the qualitative analysis, it could determine those three regions with high accuracy. After that, contrast of each region will be enhanced using a new local contrast enhancement technique called Adaptive Fuzzy Exposure Local Contrast Enhancement (AFELCE). The proposed AFELCE method is specifically designed to enhance the contrast of each region using different approaches. The proposed AFELCE technique successfully improves the contrast of 300 low-contrast and non-uniform illumination images, taken from three different databases namely standard, underwater, and microscopic human sperm images. The proposed AFELCE method qualitatively and quantitatively outperforms the state-of-the-art methods,. Qualitatively, the proposed AFELCE method has successfully enhanced the contrast of those images by producing more uniform illumination images with high contrast. Quantitatively, the proposed AFELCE method produces the highest average of Entropy (E), Measure of Enhancement (EME) and Universal Image Quality Index (UIQI) for the standard image database with values of 7.582, 42.75 and 0.94 respectively. The similar results obtained for the underwater database images, where it produces the highest average of E, EME and UIQI values with 7.124, 41.13 and 0.89 respectivley. While for the microscopic human sperm image database, it produces the highest values for E and EME with values of 7.602 and 42.51 respectively, and . This study is suitable to be applied to a real time applications. Based on the good results obtained for standard, underwater, and microscopic human sperm images, the developed system has high potential and suitable to be applied to a real time applications.