Ringing artifact assessment/reduction and low quality image interpolation
Abstract
Today, digital images have become inseparable from our daily life. They are widely used in applications such as Internet, printing, and medical imaging. The quality of digital images may be degraded during the process of capture, storage, or transmission. Therefore, quality assessment and enhancement are desired in various applications, especially for low quality images such as heavily compressed or low resolution images. In this thesis, we address two problems in these areas. In the first part, we present a no-reference method to assess and reduce a specific type of artifact in JPEG compressed images, the ringing artifact. It refers to the noisy-looking vicinity of edges that is supposed to belong to a relatively smooth object. In our method, to measure the visual impact of a ringing artifact, we compare its activity level with the activity of the object to which it belongs, as well as a local luminance-mediated just-noticeable noise level. Based on the measurement, we also propose a ringing artifact reduct on method. Our measurement results show consistency with the perceptual impact of the ringing artifacts. We also achieve improvements in ringing artifact reduction over another method in the literature. In the second part of our work, we develop an adaptive interpolation method for low quality images such as video image. Our method chooses among four different interpolation schemes: bilinear, edge-directed, peaking, or texture synthesis, according to local image context. The interpolation results show a visible improvement compared to those of traditional interpolation methods, with sharper edges, richer texture, and less noise in smooth regions.
Degree
Ph.D.
Advisors
Allebach, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.