[vlc-devel] Please help in blending routine

[R. Bernstein]

I've been trying to get a blending routine working for subtitles - CVD
and SVCD kinds of subtitles, but I believe this routine could be used to
improve the DVD subtitle rendering as well.

It's been very tough going. Very few existing routines can be used to
simplify coding.  Documentation of the formats is very sparse: some
doxygen comments; one has to look at code to get more detailed
specification information. As far as I know there are only too
blending routines that to use as a model: modules/misc/freetype.c and
modules/codec/spudec/render.c. But neither of these is really general
purpose; the former handles font bitmaps; the latter is a lacking in
terms of handling transparancy for some video formats and doesn't
handle all color components. (usually, if not exclusively, it handles
only the black/white component). Neither is what I would call easily
readable or easily modifiable.

For example, here's a portion of freetype.c that I found relevant: 

                    FT_BitmapGlyph p_glyph = p_line->pp_glyphs[ i ];
#define alpha p_vout->p_text_renderer_data->pi_gamma[ p_glyph->bitmap.buffer[ x + y * p_glyph->bitmap.width ] ]
#define pixel p_in[ ( p_line->p_glyph_pos[ i ].y + pen_y + y - p_glyph->top ) * i_pic_pitch + x + pen_x + p_line->p_glyph_pos[ i ].x + p_glyph->left ]
                    for(y = 0; y < p_glyph->bitmap.rows; y++ )
                    {
                        for( x = 0; x < p_glyph->bitmap.width; x++ )
                        {
                            pen_y--;
                            pixel = ( ( pixel * ( 255 - alpha ) ) >> 8 );
                            pen_y++; pen_x--;
                            pixel = ( ( pixel * ( 255 - alpha ) ) >> 8 );
                            pen_x += 2;
                            pixel = ( ( pixel * ( 255 - alpha ) ) >> 8 );
                            pen_y++; pen_x--;
                            pixel = ( ( pixel * ( 255 - alpha ) ) >> 8 );
                            pen_y--;
                        }
                    }
                    for(y = 0; y < p_glyph->bitmap.rows; y++ )
                    {
                        for( x = 0; x < p_glyph->bitmap.width; x++ )
                        {
                            pixel = ( ( pixel * ( 255 - alpha ) ) >> 8 ) +
                                ( 255 * alpha >> 8 );
                        }
                    }

I cite the above (which is a more correct implimentation of blending)
not because I want an explanation of the above, but to give an idea of
what one is faced with to accomplish writing a blending routine for
vlc. And as pointed out before, the above is the more *correct*
implimentation of the two. The spudec render.c routine is as obscure
if not also incomplete with it's countdown loops which are subtracted
from the end of an array to give forward access.

What I have so far is in: modules/codec/ogt/render.c. Help, thoughts,
comments, code improvements would be greatly appreciated. Thanks.

-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>