It is possible to recreate the original movie table by joining all the tables back together. First, join the associative tables to the actor/director tables. Then pivot the num column, and add the column prefixes back:

>>> actors = actor_associative.merge(actor_unique, on='actor_id') 
                              .drop('actor_id', 1) 
                              .pivot_table(index='id', 
                                           columns='num',
                                           aggfunc='first')

>>> actors.columns = actors.columns.get_level_values(0) + '_' + 
                     actors.columns.get_level_values(1).astype(str)

>>> directors = director_associative.merge(director_unique,
                                           on='director_id') 
                                    .drop('director_id', 1) 
                                    .pivot_table(index='id',
                                                 columns='num',
                                                 aggfunc='first')

>>> directors.columns = directors.columns.get_level_values(0) + '_' + 
                        directors.columns.get_level_values(1) 
                                         .astype(str)

These tables can now be joined together with movie_table:

>>> movie2 = movie_table.merge(directors.reset_index(),
                               on='id', how='left') 
                        .merge(actors.reset_index(),
                               on='id', how='left')
>>> movie.equals(movie2[movie.columns])
True