Publication Date

Spring 2023

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering

Advisor

Chang Choo

Subject Areas

Electrical engineering

Abstract

In this thesis, an image interpolation scheme using an adaptive windowed sinc, which varies in length and frequency response guided by the Laplacian operator, is proposed to improve quality and accuracy. First, an optimization is performed to minimize the squared error for each sliding window by sweeping a range of windowed sinc filters with varying filter lengths. The relationship between the windows, filter length, and edge intensity is analyzed to determine the optimal beta value and filter length. This analysis inspires an image interpolation approach that uses the Laplacian of each sliding window to choose the optimal beta and filter length. The performance of the proposed approach is compared with the optimized result, as well as traditional interpolation methods such as bilinear and bicubic, in terms of PSNR, SSIM, and a subjective visual test. The findings demonstrate the effectiveness of the proposed method in enhancing the resolution of images, contributing to a deeper understanding of adaptive windowed sinc filters and their potential applications in image interpolation.

Download

Share

COinS