Date of Award


Degree Name

Doctor of Philosophy


Electrical and Computer Engineering

First Advisor

Gupta, Lalit


This dissertation focuses on the modeling of dust noise on ordinary images and formulating enhancement and restoration strategies to improve the quality of dusty images using a sequence of image processing steps. Analyses of images acquired in dusty environments show that the images tend to have noise, blur, a small dynamic range, low contrast, diminished blue components, and high red components. This dissertation proposes that the dust noise model on ordinary images consists of the three steps. The first step contends that the atmospheric turbulence-blurring model proposed by Hufnagel and Stanley is used to generate a blur noise. The second step is that the gamma correction is used to adjust image brightness. The gamma parameter is calculated directly from the color components without prior knowledge. The third step consists of a statistical method that is used to simulate real dust images colors contrasts. This noise model is designed based on a statistical analysis of real dust images histograms. It is an adaptive model that can extract necessary parameters directly from the original images without prior knowledge in order to degrade each color component independently. As a result, the dust noise model could simulate real dust images. The second contribution of this dissertation is to develop an automatic color correction algorithm (D) that improves the quality of dusty images. This algorithm is based on the Wiener filter, luminance stretching, and a modified homomorphic filter. The Wiener filter is applied to restore image pixels and remove the blur noise. Subsequently, an image is converted from RGB to YCbCr color model. In YCbCr, the luminance is stretched to adjust the illumination of the image at the same level automatically. In the RGB color model, the modified homomorphic filter is applied to enhance the image contrast in order to get true colors. The third contribution of this dissertation is to develop statistical adaptive algorithms (S) that consists of the following: use of the Wiener filter to restore image pixels and application of contrast stretching methods that improve the contrast of an image. Every pixel value in all color components of red, green, and blue is stretched between the smallest and largest values using the same scaling function in order to preserve the correct contrast. Once this is done, the image is converted into the HSI color model in order to stretch the intensity. The intensity is adjusted to get true colors and illumination. Finally, a color balance approach is used to remove colorcast. Enhancement experiments are conducted on real and simulated dusty images. Both algorithms are evaluated using the four well-known methods: human perception, root mean square error, peak signal to noise ratio, and structural similarity index. The use of these methods ensures that the introduced algorithms are thoroughly evaluated. The results show that the introduced algorithms are quite effective in enhancing dusty images. Furthermore, the results are superior to those obtained through histogram equalization, gray world, and white patch algorithms. In addition, the complexities of the introduced algorithms are very low, making them attractive for real time-image processing.




This dissertation is only available for download to the SIUC community. Current SIUC affiliates may also access this paper off campus by searching Dissertations & Theses @ Southern Illinois University Carbondale from ProQuest. Others should contact the interlibrary loan department of your local library or contact ProQuest's Dissertation Express service.