Generic method for 2D image resizing with non-separable filters

Abstract

2D image resizing is an important issue for pixel oriented displays with variable input formats. Low-resolution pictures look bad on high-resolution screens, especially when only simple up-conversion methods like pixel and line repetition or bi-linear interpolation are used. Even when applying separable polyphase up-conversion filters, the problem of jagged lines (staircases) remains. The approach investigated for high-quality resizing by rational factors is based on pixel and line repetition with suitable post-filtering. The novel resizing method uses very simple, non-separable filters, and is suitable for all kinds of image and video material such as analog sources (PAL, NTSC), digital sources (JPEG, MPEG), low-resolution up to high-resolution images, and noisy pictures. Jagged lines can be smoothened perfectly.