Improving noise immunity of homography superposition for television signals

Abstract

The algorithm of homography superposition of the television signals was shown in this article. The algorithm allows improving noise immunity for homography parameters of television signal with textural fragments. The core of the estimation procedure of homography parameters for image superposition is the matching of the feature points. The feature points are local extremes of the pixel brightness of the image. The matching is defined by the maximum correlation coefficient between a region of the image of a neighbor area of the feature point. The log-polar system of the images provides an invariant correlation of scale and rotate are proportional offsets along the axis in this coordinate system. The noise immunity is provided due to the developing procedure of removing feature points in the textural region. This procedure leads to decreasing probability of error matching of the feature points. The numerical modeling shows high noise immunity of the developed procedure in comparing the current researches.