Dissertation > Excellent graduate degree dissertation topics show

Research of Digital Image Inpainting Technology

Author: ZuoZhenJia
Tutor: NiuXiaMu
School: Harbin Institute of Technology
Course: Computer Science and Technology
Keywords: image inpainting analytic hierarchy structure extraction structurepairing
CLC: TP391.41
Type: Master's thesis
Year: 2012
Downloads: 64
Quote: 0
Read: Download Dissertation

Abstract

Image inpainting, also known as image completion is the art of filling missing ordamaged regions of an image in a plausible way to restore its integrity and thequality of visual perception so that an ordinary observer hardly detects whetherthe image had been damaged. Image inpainting can be used in many areas, suchas protection of historical relics, visual special effects, virtual reality and so on. Ithas become an important topic in image processing and computer vision for itswide variety of application domains with broad prospects as well as theoreticalvalue. In this thesis, we focus on image inpainting and our work mainly includesthe following aspects.1) Inspired by the hierarchy of natural language’s representation and processing,we take image as a special language. And the image’s representation andcorresponding algorithms have a similar hierarchy structure. Along this idea,we classify the existing image inpainting algorithms into three levels, namely,data-based level, structure-based level and semantics-based level. Thisanalytic hierarchy process helps us to research and improve different imageinpainting algorithms.2) We improve an image inpainting algorithm based on salient structurecompletion and texture synthesis. In the salient structure extract step, theoriginal algorithm can’t set the threshold parameters automatically. To solvethis problem, we propose a salient structure extract method based onthreshold decreasing in the Canny edge detector. We have also modified theincomplete salient structure paring distance function by adding to which aconnect-cost term and it improves the final quality of structure completion.3) We design and implement an application-oriented image inpainting systemintegrating a variety of different image inpainting algorithms. Currently,image inpainting algorithms are generally faced with the problem of poorrobustness. We propose a simple scheme which may choose the mostappropriate one from the image inpainting algorithms library without anyhuman interventions.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile