From 37644a0cee9695b2d4db0ed2ed29c85da676b7b1 Mon Sep 17 00:00:00 2001 From: lorenm Date: Tue, 28 Dec 2004 01:53:17 +0000 Subject: new x264 options: b8x8mv, direct_pred changed explanation: 4x4mv changed defaults: subq, pb_factor git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@14260 b3059339-0415-0410-9bf9-f77b7e298cf2 --- DOCS/man/en/mplayer.1 | 50 +++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 41 insertions(+), 9 deletions(-) (limited to 'DOCS') diff --git a/DOCS/man/en/mplayer.1 b/DOCS/man/en/mplayer.1 index bef6773703..a41b21e6de 100644 --- a/DOCS/man/en/mplayer.1 +++ b/DOCS/man/en/mplayer.1 @@ -7193,7 +7193,7 @@ quantizer factor between I- and P-frames (default: 1.4) . .TP .B pb_factor= -quantizer factor between P- and B-frames (default: 1.4) +quantizer factor between P- and B-frames (default: 1.3) . .TP .B pass=<1\-3> @@ -7262,14 +7262,46 @@ Lower values allow the quantizer value to jump around more, higher values force it to vary more smoothly. . .TP +.B direct_pred=<0\-2> +Determines the type of motion prediction used for direct macroblocks +in B-frames. +.PD 0 +.RSs +.IPs 0 +None: direct macroblocks are not used. +.IPs 1 +Temporal: motion vectors are interpolated from the following P-frame. +.IPs 2 +Spatial: motion vectors are extrapolated from neighboring blocks. +.RE +.PD 1 +Spatial and temporal are approximately the same speed, and neither is +strictly better than the other. +The best choice depends on the movie. +direct_pred=0 is usually both slower and lower quality. +. +.TP +.B (no)b8x8mv +Use macroblock types b16x8, b8x16, b8x8 (default: enabled). +Requires B-frames. +See 4x4mv for details. +. +.TP .B (no)4x4mv -Use all available interframe macroblock types (i16x16, i4x4, p16x16, -p16x8, p8x16, p8x8, p8x4, p4x8, p4x4, skip) +Use macroblock types p8x4, p4x8, p4x4 (default: disabled). +.br +The available interframe macroblock types are i16x16, i4x4, p16x16, p16x8, +p8x16, p8x8, p8x4, p4x8, p4x4, b16x16, b16x8, b8x16, b8x8, skip, direct. +i16x16, i4x4, p16x16-8x8, b16x16, and skip are always used. +.br The idea is to find the type and size that best describe a certain area -of the picture, i.e.\& very effective for Anime, which usually contains -large areas of the same color (default: i16x16, i4x4, p16x16-8x8, skip). -Depending on the source material, it can improve or degrade quality, use -it with care. +of the picture. +For example, a global pan is better represented by 16x16 blocks, while +small moving objects are better represented by smaller blocks. +.br +4x4mv is recommended only with subq >= 3. +Depending on the source material, it can improve or degrade +quality, use it with care. . .TP .B subq=<0\-5> @@ -7280,11 +7312,11 @@ It can improve quality significantly. .RSs 0: halfpel only .br -1: 1 iteration of qpel on the winning candidate (default) +1: 1 iteration of qpel on the winning candidate .br 2: 2 iterations of qpel on the winning candidate .br -3: halfpel on all MB types, qpel on the winner +3: halfpel on all MB types, qpel on the winner (default) .br 4: qpel on all .br -- cgit v1.2.3