Performance comparison on algorithms of image resizing

• Main Author: Jiang, Yuwei
• Other Authors: Bi Guoan
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: http://hdl.handle.net/10356/54433
• Institution: Nanyang Technological University

Image resizing is very common in daily life. They are used in a broad range of applications. For example, image search engines (e.g., Google and Bing) need to display the resized thumbnail images for the search results. Medical images are resized to assist doctors for better medical check-up. Smart mobile phone users can zoom in or zoom out an image with two fingers for better browsing experience. In computer vision, researchers always resize the images to a fixed size for better analyzing. To date, many image resizing algorithms developed, from traditional image resizing algorithms using interpolation to content-aware image resizing algorithms using some advanced techniques. There is thus a need to compare the performance of these image resizing algorithms in order to understand their characteristics, and to guide the selection of proper image resizing algorithm for a specific application. In this report, we review three popular algorithms, namely nearest neighbor, bilinear and bicubic interpolation and conduct experimental study on them in order to compare their performances in different situations. An interactive user interface is also developed for non-programming users to visually compare the differences of using different image resizing algorithms. Basics in image processing are introduced in this paper to provide foundation knowledge for image resizing. The performance is evaluated in terms of mean square error. Experiments are conducted to test the functionality of user interface. Moreover, experiments on performance comparison of resizing algorithms are conducted based on 1) fixed image size and 2) sampling frequency.