Measurement of Human Sensitivity across the Vertical-Temporal Video Spectrum for Interlacing Filter Specification

Abstract

Good quality conversion from progressive to interlaced video is highly relevant to today's broadcast systems, in which interlaced content is still common. The interlacing process is a form of down-sampling, and hence requires an anti-alias filter. For best results the anti-alias filter should be matched to the reconstruction filter, which is comprised of the display and the human visual system. Additionally, it must meet the technical requirements of the down sampling process. In this paper we present a novel method of measuring the combined response to interlacing artefacts that is simple and powerful. We use the results to derive an optimal anti-alias filter template, using a new region-growing technique that is specifically designed to match the measured response whilst keeping to the technical constraints of an interlaced sampling structure. Our results provide support for an existing, heuristically-defined filter, and show that the same filter could be used for a range of viewing distances.