Optimized interpolation of discrete picture signals
Kouji Kinuhata · Electronics and Communications in Japan (Part I Communications) · 1980
Abstract This paper describes the optimization of linear interpolation of picture elements in terms of least mean square error for pictures of a square array of elements. Two kinds of models are assumed for autocorrelation function m, n which is defined as the correlation of two elements separated m picture elements in a horizontal direction and n picture elements in a vertical direction. One is a homogeneous model in which m, n is given by the equation m, n In the interpolation of the picture element at the point not sampled, the optimum interpolation can be achieved by the linear combination of the four picture elements nearest to the point to be interpolated. This is strictly true in the case of an inhomogeneous correlation function and approximately true in the case of a homogeneous correlation function. On the other hand, in the interpolation of the picture element at the sampling point, the interpolation can be optimized by the linear combination of eight picture elements around the point to be interpolated. This is strictly true in the case of an inhomogeneous correlation function and approximately true in the case of a homogeneous correlation function. Using the above results, interpolation techniques in “standard conversion in broadcasting television signals” and “system conversion from video phone signal to broadcasting television signal” are reviewed, and the pictures processed with the interpolation techniques are evaluated.