The biproportional filter was created to analyze structural change between twoinput-output matrices by removing the effect of differential growth of sectors withoutpredetermining if the model is demand or supply-driven, but with the disadvantage thatprojecting a first matrix on a second is not the same thing than projecting the second matrix on the first. Here two alternative methods are proposed which has not this last drawback, with the additional advantage for the biproportional bimarkovian filter that effects of sector size are also removed. Methods are compared with an application for France for 1980 and 1996.