Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Automatic Alignment of Video Chrominance

IP.com Disclosure Number: IPCOM000108679D
Original Publication Date: 1992-Jun-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 2 page(s) / 95K

Publishing Venue

IBM

Related People

Edgar, AD: AUTHOR

Abstract

All forms of consumer video segregate the luminance and chrominance with different bandwidths at each stage. Because of this, color and luminance are usually out of time alignment, producing color ghosts and smear. The algorithm of this article measures misalignment from natural images. From this measurement, time alignment and associated phase linearity are restored between the luminance and chrominance video components.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Automatic Alignment of Video Chrominance

       All forms of consumer video segregate the luminance and
chrominance with different bandwidths at each stage. Because of this,
color and luminance are usually out of time alignment, producing
color ghosts and smear.  The algorithm of this article measures
misalignment from natural images. From this measurement, time
alignment and associated phase linearity are restored between the
luminance and chrominance video components.

      The present invention brings the color and brightness signals
back into phase alignment in order to utilize the full available
color bandwidth.  Time-misalignment is a phase error, one of several
errors that will be corrected by aligning phase.  The misalignment
data will be extracted from the real images, without a special test
chart, and is thus useful for general application.

      First, for simplicity, the image is converted back to a
one-dimensional scan signal, because the induced errors were, in
fact, introduced in just the one time dimension.

      Second, the signal is split into its three color components, Y,
I, and Q for NTSC.  Each of these components represents the same
image processes, reproducing the same edges and details, so there is
a strong correlation between them in real images.  This assumption
that the two color components arise from the same physical structure
as the luminance component, and therefore edges correlate, is
critical to the operation of this algorithm.

      Third, each of these three signals is split into a series of
bandpass signals such that when the bandpass signals are re-added,
the original signal is obtained. These bands are one or one-half
octave in width.  The signal may be decomposed into bands using a
pyramid structure.

      Fourth, for each frequency band of each of the two color
signals, perform a cross-correlation with the same frequency band of
the luminance signal.  The cross-correlation will peak off center by
an amount by which the signals are miscorrelated in time at that
particular frequency.  There is a subtle but critical twist.
Although the phases between Y and I and between Y and Q should be
equal when properly aligned, their magnitudes are as likely to be
exact opp...