The invention relates to a method for encoding video signals corresponding to a sequence of frames constituted either from video-type images or film-type images. These film-type images are obtained by means of the 3:2 pull-down technique involving a subdivision of each film frame into two fields and a duplication of two fields within each group of four subdivided frames. The encoding step of the signals is preceded by a preprocessing step that includes thresholding the eight absolute values of the successive pixel-to-pixel differences between fields of the same parity. Measuring the density of "1" (resp. "0") per block of mn pixels of the field structure leading to mark each pixel as 1 or 0. Detecting in any group of ten successive fields within the preprocessed signals thus obtained, the film pattern formed by the duplicated fields. Switching the preprocessed signals from a first encoding mode to a second one, when the film pattern is detected, or from the second one to the first one, when it is not detected.