vpx/vp9/encoder/vp9_picklpf.c

263 lines
7.8 KiB
C
Raw Normal View History

2010-05-18 17:58:33 +02:00
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
2010-05-18 17:58:33 +02:00
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
2010-05-18 17:58:33 +02:00
*/
#include <assert.h>
#include <limits.h>
#include "vp9/common/vp9_onyxc_int.h"
#include "vp9/encoder/vp9_onyx_int.h"
#include "vp9/encoder/vp9_picklpf.h"
#include "vp9/encoder/vp9_quantize.h"
2010-05-18 17:58:33 +02:00
#include "vpx_mem/vpx_mem.h"
#include "vpx_scale/vpx_scale.h"
#include "vp9/common/vp9_alloccommon.h"
#include "vp9/common/vp9_loopfilter.h"
#include "./vpx_scale_rtcd.h"
2010-05-18 17:58:33 +02:00
void vp9_yv12_copy_partial_frame_c(YV12_BUFFER_CONFIG *src_ybc,
YV12_BUFFER_CONFIG *dst_ybc, int Fraction) {
uint8_t *src_y, *dst_y;
int yheight;
int ystride;
int yoffset;
int linestocopy;
2010-05-18 17:58:33 +02:00
assert(src_ybc->y_stride == dst_ybc->y_stride);
yheight = src_ybc->y_height;
ystride = src_ybc->y_stride;
2010-05-18 17:58:33 +02:00
linestocopy = (yheight >> (Fraction + 4));
2010-05-18 17:58:33 +02:00
if (linestocopy < 1)
linestocopy = 1;
2010-05-18 17:58:33 +02:00
linestocopy <<= 4;
2010-05-18 17:58:33 +02:00
yoffset = ystride * ((yheight >> 5) * 16 - 8);
src_y = src_ybc->y_buffer + yoffset;
dst_y = dst_ybc->y_buffer + yoffset;
2010-05-18 17:58:33 +02:00
vpx_memcpy(dst_y, src_y, ystride * (linestocopy + 16));
2010-05-18 17:58:33 +02:00
}
static int calc_partial_ssl_err(YV12_BUFFER_CONFIG *source,
YV12_BUFFER_CONFIG *dest, int Fraction) {
int i, j;
int Total = 0;
int srcoffset, dstoffset;
uint8_t *src = source->y_buffer;
uint8_t *dst = dest->y_buffer;
2010-05-18 17:58:33 +02:00
int linestocopy = (source->y_height >> (Fraction + 4));
2010-05-18 17:58:33 +02:00
if (linestocopy < 1)
linestocopy = 1;
2010-05-18 17:58:33 +02:00
linestocopy <<= 4;
2010-05-18 17:58:33 +02:00
srcoffset = source->y_stride * (dest->y_height >> 5) * 16;
dstoffset = dest->y_stride * (dest->y_height >> 5) * 16;
2010-05-18 17:58:33 +02:00
src += srcoffset;
dst += dstoffset;
2010-05-18 17:58:33 +02:00
// Loop through the Y plane raw and reconstruction data summing (square differences)
for (i = 0; i < linestocopy; i += 16) {
for (j = 0; j < source->y_width; j += 16) {
unsigned int sse;
Total += vp9_mse16x16(src + j, source->y_stride, dst + j, dest->y_stride,
&sse);
2010-05-18 17:58:33 +02:00
}
src += 16 * source->y_stride;
dst += 16 * dest->y_stride;
}
return Total;
2010-05-18 17:58:33 +02:00
}
// Enforce a minimum filter level based upon baseline Q
static int get_min_filter_level(VP9_COMP *cpi, int base_qindex) {
int min_filter_level;
/*int q = (int) vp9_convert_qindex_to_q(base_qindex);
if (cpi->source_alt_ref_active && cpi->common.refresh_golden_frame && !cpi->common.refresh_alt_ref_frame)
min_filter_level = 0;
else
{
if (q <= 10)
min_filter_level = 0;
else if (q <= 64)
min_filter_level = 1;
else
min_filter_level = (q >> 6);
}
*/
min_filter_level = 0;
return min_filter_level;
2010-05-18 17:58:33 +02:00
}
// Enforce a maximum filter level based upon baseline Q
static int get_max_filter_level(VP9_COMP *cpi, int base_qindex) {
// PGW August 2006: Highest filter values almost always a bad idea
2010-05-18 17:58:33 +02:00
// jbb chg: 20100118 - not so any more with this overquant stuff allow high values
// with lots of intra coming in.
int max_filter_level = MAX_LOOP_FILTER;// * 3 / 4;
(void)base_qindex;
2010-05-18 17:58:33 +02:00
if (cpi->twopass.section_intra_rating > 8)
max_filter_level = MAX_LOOP_FILTER * 3 / 4;
2010-05-18 17:58:33 +02:00
return max_filter_level;
2010-05-18 17:58:33 +02:00
}
// Stub function for now Alt LF not used
void vp9_set_alt_lf_level(VP9_COMP *cpi, int filt_val) {
WebM Experimental Codec Branch Snapshot This is a code snapshot of experimental work currently ongoing for a next-generation codec. The codebase has been cut down considerably from the libvpx baseline. For example, we are currently only supporting VBR 2-pass rate control and have removed most of the code relating to coding speed, threading, error resilience, partitions and various other features. This is in part to make the codebase easier to work on and experiment with, but also because we want to have an open discussion about how the bitstream will be structured and partitioned and not have that conversation constrained by past work. Our basic working pattern has been to initially encapsulate experiments using configure options linked to #IF CONFIG_XXX statements in the code. Once experiments have matured and we are reasonably happy that they give benefit and can be merged without breaking other experiments, we remove the conditional compile statements and merge them in. Current changes include: * Temporal coding experiment for segments (though still only 4 max, it will likely be increased). * Segment feature experiment - to allow various bits of information to be coded at the segment level. Features tested so far include mode and reference frame information, limiting end of block offset and transform size, alongside Q and loop filter parameters, but this set is very fluid. * Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used in MBs using 16x16 prediction modes within inter frames. * Compound prediction (combination of signals from existing predictors to create a new predictor). * 8 tap interpolation filters and 1/8th pel motion vectors. * Loop filter modifications. * Various entropy modifications and changes to how entropy contexts and updates are handled. * Extended quantizer range matched to transform precision improvements. There are also ongoing further experiments that we hope to merge in the near future: For example, coding of motion and other aspects of the prediction signal to better support larger image formats, use of larger block sizes (e.g. 32x32 and up) and lossless non-transform based coding options (especially for key frames). It is our hope that we will be able to make regular updates and we will warmly welcome community contributions. Please be warned that, at this stage, the codebase is currently slower than VP8 stable branch as most new code has not been optimized, and even the 'C' has been deliberately written to be simple and obvious, not fast. The following graphs have the initial test results, numbers in the tables measure the compression improvement in terms of percentage. The build has the following optional experiments configured: --enable-experimental --enable-enhanced_interp --enable-uvintra --enable-high_precision_mv --enable-sixteenth_subpel_uv CIF Size clips: http://getwebm.org/tmp/cif/ HD size clips: http://getwebm.org/tmp/hd/ (stable_20120309 represents encoding results of WebM master branch build as of commit#7a15907) They were encoded using the following encode parameters: --good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63 --end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999 --kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50 --minsection-pct=0 --maxsection-pct=800 --sharpness=0 --arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF) --arnr-type=3 Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
}
2010-05-18 17:58:33 +02:00
void vp9_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP9_COMP *cpi) {
VP9_COMMON *cm = &cpi->common;
2010-05-18 17:58:33 +02:00
int best_err = 0;
int filt_err = 0;
int min_filter_level = get_min_filter_level(cpi, cm->base_qindex);
int max_filter_level = get_max_filter_level(cpi, cm->base_qindex);
2010-05-18 17:58:33 +02:00
int filter_step;
int filt_high = 0;
int filt_mid = cm->filter_level; // Start search at previous frame filter level
int filt_low = 0;
int filt_best;
int filt_direction = 0;
2010-05-18 17:58:33 +02:00
int Bias = 0; // Bias against raising loop filter and in favour of lowering it
2010-05-18 17:58:33 +02:00
// Make a copy of the unfiltered / processed recon buffer
vp8_yv12_copy_y(cm->frame_to_show, &cpi->last_frame_uf);
2010-05-18 17:58:33 +02:00
if (cm->frame_type == KEY_FRAME)
cm->sharpness_level = 0;
else
cm->sharpness_level = cpi->oxcf.Sharpness;
2010-05-18 17:58:33 +02:00
// Start the search at the previous frame filter level unless it is now out of range.
filt_mid = cm->filter_level;
2010-05-18 17:58:33 +02:00
if (filt_mid < min_filter_level)
filt_mid = min_filter_level;
else if (filt_mid > max_filter_level)
filt_mid = max_filter_level;
2010-05-18 17:58:33 +02:00
// Define the initial step size
filter_step = (filt_mid < 16) ? 4 : filt_mid / 4;
2010-05-18 17:58:33 +02:00
// Get baseline error score
vp9_set_alt_lf_level(cpi, filt_mid);
vp9_loop_filter_frame(cm, &cpi->mb.e_mbd, filt_mid, 1, 0);
2010-05-18 17:58:33 +02:00
best_err = vp9_calc_ss_err(sd, cm->frame_to_show);
filt_best = filt_mid;
2010-05-18 17:58:33 +02:00
// Re-instate the unfiltered frame
vp8_yv12_copy_y(&cpi->last_frame_uf, cm->frame_to_show);
2010-05-18 17:58:33 +02:00
while (filter_step > 0) {
Bias = (best_err >> (15 - (filt_mid / 8))) * filter_step; // PGW change 12/12/06 for small images
2010-05-18 17:58:33 +02:00
// jbb chg: 20100118 - in sections with lots of new material coming in don't bias as much to a low filter value
if (cpi->twopass.section_intra_rating < 20)
Bias = Bias * cpi->twopass.section_intra_rating / 20;
2010-05-18 17:58:33 +02:00
// yx, bias less for large block size
if (cpi->common.txfm_mode != ONLY_4X4)
Bias >>= 1;
WebM Experimental Codec Branch Snapshot This is a code snapshot of experimental work currently ongoing for a next-generation codec. The codebase has been cut down considerably from the libvpx baseline. For example, we are currently only supporting VBR 2-pass rate control and have removed most of the code relating to coding speed, threading, error resilience, partitions and various other features. This is in part to make the codebase easier to work on and experiment with, but also because we want to have an open discussion about how the bitstream will be structured and partitioned and not have that conversation constrained by past work. Our basic working pattern has been to initially encapsulate experiments using configure options linked to #IF CONFIG_XXX statements in the code. Once experiments have matured and we are reasonably happy that they give benefit and can be merged without breaking other experiments, we remove the conditional compile statements and merge them in. Current changes include: * Temporal coding experiment for segments (though still only 4 max, it will likely be increased). * Segment feature experiment - to allow various bits of information to be coded at the segment level. Features tested so far include mode and reference frame information, limiting end of block offset and transform size, alongside Q and loop filter parameters, but this set is very fluid. * Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used in MBs using 16x16 prediction modes within inter frames. * Compound prediction (combination of signals from existing predictors to create a new predictor). * 8 tap interpolation filters and 1/8th pel motion vectors. * Loop filter modifications. * Various entropy modifications and changes to how entropy contexts and updates are handled. * Extended quantizer range matched to transform precision improvements. There are also ongoing further experiments that we hope to merge in the near future: For example, coding of motion and other aspects of the prediction signal to better support larger image formats, use of larger block sizes (e.g. 32x32 and up) and lossless non-transform based coding options (especially for key frames). It is our hope that we will be able to make regular updates and we will warmly welcome community contributions. Please be warned that, at this stage, the codebase is currently slower than VP8 stable branch as most new code has not been optimized, and even the 'C' has been deliberately written to be simple and obvious, not fast. The following graphs have the initial test results, numbers in the tables measure the compression improvement in terms of percentage. The build has the following optional experiments configured: --enable-experimental --enable-enhanced_interp --enable-uvintra --enable-high_precision_mv --enable-sixteenth_subpel_uv CIF Size clips: http://getwebm.org/tmp/cif/ HD size clips: http://getwebm.org/tmp/hd/ (stable_20120309 represents encoding results of WebM master branch build as of commit#7a15907) They were encoded using the following encode parameters: --good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63 --end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999 --kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50 --minsection-pct=0 --maxsection-pct=800 --sharpness=0 --arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF) --arnr-type=3 Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
filt_high = ((filt_mid + filter_step) > max_filter_level) ? max_filter_level : (filt_mid + filter_step);
filt_low = ((filt_mid - filter_step) < min_filter_level) ? min_filter_level : (filt_mid - filter_step);
2010-05-18 17:58:33 +02:00
if ((filt_direction <= 0) && (filt_low != filt_mid)) {
// Get Low filter error score
vp9_set_alt_lf_level(cpi, filt_low);
vp9_loop_filter_frame(cm, &cpi->mb.e_mbd, filt_low, 1, 0);
2010-05-18 17:58:33 +02:00
filt_err = vp9_calc_ss_err(sd, cm->frame_to_show);
2010-05-18 17:58:33 +02:00
// Re-instate the unfiltered frame
vp8_yv12_copy_y(&cpi->last_frame_uf, cm->frame_to_show);
2010-05-18 17:58:33 +02:00
// If value is close to the best so far then bias towards a lower loop filter value.
if ((filt_err - Bias) < best_err) {
// Was it actually better than the previous best?
if (filt_err < best_err)
best_err = filt_err;
2010-05-18 17:58:33 +02:00
filt_best = filt_low;
}
}
2010-05-18 17:58:33 +02:00
// Now look at filt_high
if ((filt_direction >= 0) && (filt_high != filt_mid)) {
vp9_set_alt_lf_level(cpi, filt_high);
vp9_loop_filter_frame(cm, &cpi->mb.e_mbd, filt_high, 1, 0);
2010-05-18 17:58:33 +02:00
filt_err = vp9_calc_ss_err(sd, cm->frame_to_show);
2010-05-18 17:58:33 +02:00
// Re-instate the unfiltered frame
vp8_yv12_copy_y(&cpi->last_frame_uf, cm->frame_to_show);
// Was it better than the previous best?
if (filt_err < (best_err - Bias)) {
best_err = filt_err;
filt_best = filt_high;
}
}
// Half the step distance if the best filter value was the same as last time
if (filt_best == filt_mid) {
filter_step = filter_step / 2;
filt_direction = 0;
} else {
filt_direction = (filt_best < filt_mid) ? -1 : 1;
filt_mid = filt_best;
2010-05-18 17:58:33 +02:00
}
}
2010-05-18 17:58:33 +02:00
cm->filter_level = filt_best;
#if CONFIG_LOOP_DERING
/* Decide whether to turn on deringing filter */
{ // NOLINT
int best_dering = 0;
int this_dering;
int last_err_diff = INT_MAX;
for (this_dering = 1; this_dering <= 16; this_dering++) {
vp9_set_alt_lf_level(cpi, filt_best);
vp9_loop_filter_frame(cm, &cpi->mb.e_mbd, filt_high, 1, this_dering);
filt_err = vp9_calc_ss_err(sd, cm->frame_to_show);
vp8_yv12_copy_y(&cpi->last_frame_uf, cm->frame_to_show);
if (filt_err < best_err) {
best_err = filt_err;
best_dering = this_dering;
last_err_diff = INT_MAX;
} else {
if (filt_err - best_err > last_err_diff)
break;
last_err_diff = filt_err - best_err;
}
}
cm->dering_enabled = best_dering;
}
#endif
2010-05-18 17:58:33 +02:00
}