Changed last reference to the matches dataframe to the subpixel dataframe.

                      The maximum (positive) value that a pixel can shift in the y direction

                      without being considered an outlier

        """

        for column, default in {'join_idx': -1, 'x_offset': 0, 'y_offset': 0, 'correlation': 0, 'reference': -1}.items():

        for column, default in {'x_offset': 0, 'y_offset': 0, 'correlation': 0, 'reference': -1}.items():

            if column not in self.subpixel.columns:

                self.subpixel[column] = default

            d_search = sp.clip_roi(d_img, d_keypoint, search_size)

            try:

                x_offset, y_offset, strength = sp.subpixel_offset(s_template, d_search, **kwargs)

                self.subpixel.loc[i, ('join_idx', 'x_offset', 'y_offset', 'correlation', 'reference')]= [idx, x_offset, y_offset, strength, source_image]

                self.subpixel.loc[idx, ('x_offset', 'y_offset', 'correlation', 'reference')]= [x_offset, y_offset, strength, source_image]

            except:

                warnings.warn('Template-Search size mismatch, failing for this correspondence point.')

        # Compute the mask for the point shifts that are too large

        query_string = 'x_offset <= -{0} or x_offset >= {0} or y_offset <= -{1} or y_offset >= {1}'.format(max_x_shift,max_y_shift)

        sp_shift_outliers = self.subpixel.query(query_string)

        shift_mask = pd.Series(True, index=self.matches.index)

        shift_mask = pd.Series(True, index=self.subpixel.index)

        shift_mask.loc[sp_shift_outliers.index] = False

        # Generate the composite mask and write the masks to the mask data structure