[FFmpeg-soc] [soc]: r5533 - in indeo5: TODO description

[diego]

Author: diego
Date: Sun Dec 13 12:21:05 2009
New Revision: 5533

Log:
cosmetics: Break overly long lines.

Modified:
   indeo5/TODO
   indeo5/description

Modified: indeo5/TODO
==============================================================================
--- indeo5/TODO	Sun Dec 13 05:24:02 2009	(r5532)
+++ indeo5/TODO	Sun Dec 13 12:21:05 2009	(r5533)
@@ -1,5 +1,9 @@
 Please read my comments named "description" BEFORE doing anything! Thanks!
 
-Some small steps needed to be completed in order to make FFmpeg able to decode indeo5 videos:
+Some small steps needed to be completed in order to make FFmpeg able to
+decode indeo5 videos:
 
-* split the 5/3 wavelet synthesis filter out of the snow source (see snow.c:spatial_compose53i) and adapt it to indeo5. The indeo5's high-pass coefficients must be upscaled by -2 in this case before those can be passed to that filter.
+* split the 5/3 wavelet synthesis filter out of the snow source (see
+  snow.c:spatial_compose53i) and adapt it to indeo5. The indeo5's
+  high-pass coefficients must be upscaled by -2 in this case before
+  those can be passed to that filter.

Modified: indeo5/description
==============================================================================
--- indeo5/description	Sun Dec 13 05:24:02 2009	(r5532)
+++ indeo5/description	Sun Dec 13 12:21:05 2009	(r5533)
@@ -6,20 +6,36 @@ Description of the files:
 Indeo5 docs:
 -------------
 
-There is a basic indeo5 documentation being currently in progress. It can be found here:
+There is a basic indeo5 documentation being currently in progress. It
+can be found here:
 http://wiki.multimedia.cx/index.php?title=Indeo_5
 
 Please be aware of the following important design specifics of the indeo5 codec:
 
-1. Indeo5 operates on the pixels in the range [-128...127]. It was made in order to make the values compatible with the slant transform which requires both negative and positive numbers. The conversions look like this:
+1. Indeo5 operates on the pixels in the range [-128...127]. It was made in
+   order to make the values compatible with the slant transform which requires
+   both negative and positive numbers. The conversions look like this:
 
         encoder -> internal_pixel = pixel - 128;
         decoder -> pixel = clip((internal_pixel + 128), 0, 255);
 
-2. The 5/3 wavelet filter used for splitting the signal into several bands requires at least 10-bit precision of the internal pixel values. Therefore my decoder operates completely on the 16-bit pixels by using the IDWTELEM data type. Only at the last step (output a plane) those pixels will be clipped to fit the 8-bit output values. This design requires special motion compensation routines operating on the IDWTELEM data type. The internal motion compensation routines (put_pixels stuff) cannot be used in this case!!!
+2. The 5/3 wavelet filter used for splitting the signal into several bands
+   requires at least 10-bit precision of the internal pixel values.
+   Therefore my decoder operates completely on the 16-bit pixels by using the
+   IDWTELEM data type. Only at the last step (output a plane) those pixels
+   will be clipped to fit the 8-bit output values. This design requires
+   special motion compensation routines operating on the IDWTELEM data type.
+   The internal motion compensation routines (put_pixels stuff) cannot be
+   used in this case!!!
 
-3. The 5/3 wavelet filter of indeo5 differs from the standard one in that it scales the high-pass coefficients by the value of -1/2. It should be possible to reuse the Snow's code (see snow.c:spatial_compose53i) which implements the same wavelet but upscale the indeo5 coefficients by -2 before calling this function. Maybe it will be possible to integrate the upscale step directly into this function...
+3. The 5/3 wavelet filter of indeo5 differs from the standard one in that it
+   scales the high-pass coefficients by the value of -1/2. It should be
+   possible to reuse the Snow's code (see snow.c:spatial_compose53i) which
+   implements the same wavelet but upscale the indeo5 coefficients by -2
+   before calling this function. Maybe it will be possible to integrate the
+   upscale step directly into this function...
 
-4. Please work with debugging features (see "IVI_DEBUG" in my code) enabled in order to ensure you don't break anything!
+4. Please work with debugging features (see "IVI_DEBUG" in my code)
+   enabled in order to ensure you don't break anything!
 
 For the further description please read the source or ask me...