A Novel Super Resolution Algorithm Based on Wavelet Transform and Linear Regression

  • 蘇 文伶

Student thesis: Master's Thesis


The main objective of image super resolution technology is to reconstruct a high-resolution image from a low-resolution image and make the high-resolution image become clear and natural In this Thesis a new wavelet based super resolution algorithm is proposed and it is based on the idea that the low-resolution image is the low frequency subband of a higher resolution image and the high frequency subbands are estimated to reconstruct the high-resolution image To reduce the drawbacks of traditional wavelet based super resolution algorithms such as jaggy artifacts and ringing artifacts an effective directional interpolation is adopted to accurately preserve the edges Furthermore simple linear regression is used in conjunction with the relationship of a pair of low-resolution and high-resolution images to build a reconstruction model The coefficients in high frequency subbands are estimated by the reconstruction models and a high-resolution image is generated From the experimental results it is clear that the proposed algorithm provides better performance
Date of Award2016 Aug 9
Original languageEnglish
SupervisorShen-Chuan Tai (Supervisor)

Cite this