Abstract: A motion estimation for a picture element (pixel) (P1) of a television picture (TP) is performed at pixel level by comparison of the picture signal value of the present pixel (P1) with that of a corresponding pixel (P1) in a previous and/or subsequent television picture and with that of a plurality of surrounding pixels (P2', P3', P4', P5') in the present picture. Dependent on whether a threshold value is exceeded, it is decided whether there is no motion or whether there is motion. In the comparison with the surrounding pixels, an OR function is present for the detection "no motion". It is assumed that a first and a second comparison result "no motion" associated with the pixel-pixel comparison and with the pixel-group comparison are stored for pixels (P) of the television picture (TP). Subsequently, the first and second comparison results in a sub-region (MA1) associated with the present pixel (P1) with a matrix of pixels (P1 . . .