Please use this identifier to cite or link to this item:
|Title:||Optimization Approaches to Image Enhancement|
|Department:||Electrical and Computer Engineering|
|Abstract:||To human viewers, sharp edges and rich details in an image are often interpreted as high perceptual quality. Due to various types of degradation during the acquisition of an image and the limitations of display devices and human visual system, some information that exists in the acquired image can be difficult to distinguish when the image is displayed directly, affecting its perceptual quality. Many image enhancement techniques have been proposed to fully utilize the dynamic range of the image data and reproduce a visually more appealing and informative image. In this dissertation, we present two image enhancement techniques. The first is a global approach that utilizes advanced image statistics and finds the best compromise among the factors that affect image quality; the second is a local approach exploiting the fact that the maximum discrimination power of human vision system can only be achieved in a relatively small locality of an image. These two approaches produce visually pleasing results consistently over a wide range of images. Besides the various types of artifacts, another practical problem affecting the perceptual quality of an enhanced image is the compression noise. Due to the low pass nature of image compression, the high-frequency components of a compressed image with sharp edges often carry large compression error, which can be amplified by image enhancement operator deteriorating the perceptive quality of enhanced image. By incorporating the non-linear DCT quantization mechanism into the formulation for image enhancement, we propose a new sparsity-based convex programming approach for joint quantization noise removal and enhancement. Experimental results demonstrate significant performance gains of the new approach over existing enhancement methods.|
|Appears in Collections:||Open Access Dissertations and Theses|
Files in This Item:
|shu_xiao_201601_phd.pdf||11.2 MB||Adobe PDF||View/Open|
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.