mirror of
https://github.com/ultravideo/uvg266.git
synced 2024-11-24 02:24:07 +00:00
Implement fast distortion estimation for sao.
Add function for blitting pixels from one buffer to another. Several commits have been squashed to this one.
This commit is contained in:
Ari Koivula
parent
e9819cbb5e
commit
7bd0902727
234
src/encoder.c
234
src/encoder.c
|
|
@ -772,27 +772,208 @@ void encode_slice_header(encoder_control* encoder)
|
|||
|
||||
|
||||
// TODO: move somewhere else (sao.h?)
|
||||
#define SAO_TYPE_NONE 0
|
||||
#define SAO_TYPE_EDGE 1
|
||||
#define SAO_TYPE_BAND 2
|
||||
#define Y_INDEX 0
|
||||
#define U_INDEX 1
|
||||
#define V_INDEX 2
|
||||
#define YUV_INDEX_END 3
|
||||
|
||||
#define NUM_COLORS 3
|
||||
#define NUM_SAO_OFFSETS 4
|
||||
|
||||
typedef enum { COLOR_Y = 0, COLOR_U = 1, COLOR_V = 2 } color_index;
|
||||
typedef enum { COLOR_Y = 0, COLOR_U = 1, COLOR_V = 2, NUM_COLORS } color_index;
|
||||
typedef enum { SAO_TYPE_NONE = 0, SAO_TYPE_BAND, SAO_TYPE_EDGE } sao_type;
|
||||
typedef enum { SAO_EO0 = 0, SAO_EO1, SAO_EO2, SAO_EO3, SAO_NUM_EO } sao_eo_class;
|
||||
typedef enum { SAO_EO_CAT0 = 0, SAO_EO_CAT1, SAO_EO_CAT2, SAO_EO_CAT3, SAO_EO_CAT4, NUM_SAO_EDGE_CATEGORIES } sao_eo_cat;
|
||||
|
||||
typedef struct {
|
||||
int type;
|
||||
sao_type type;
|
||||
sao_eo_class eo_class;
|
||||
int ddistortion;
|
||||
int merge_left_flag;
|
||||
int merge_up_flag;
|
||||
int offsets[NUM_SAO_OFFSETS];
|
||||
int eo_class;
|
||||
} sao_info;
|
||||
|
||||
//#define SIGN3(x) ((x) > 0) ? +1 : ((x) == 0 ? 0 : -1)
|
||||
#define SIGN3(x) (((x) > 0) - ((x) < 0))
|
||||
#define NUM_SAO_EDGE_DIRS 4;
|
||||
|
||||
typedef struct {
|
||||
int x;
|
||||
int y;
|
||||
} vector2d;
|
||||
|
||||
// Offsets of a and b in relation to c.
|
||||
// dir_offset[dir][a or b]
|
||||
// | | a | a | a |
|
||||
// | a c b | c | c | c |
|
||||
// | | b | b | b |
|
||||
static const vector2d g_sao_edge_offsets[4][2] = {
|
||||
{ { 0, -1 }, { 0, 1 } },
|
||||
{ { -1, 0 }, { 1, 0 } },
|
||||
{ { -1, -1 }, { 1, 1 } },
|
||||
{ { -1, 1 }, { 1, -1 } }
|
||||
};
|
||||
// Mapping of edge_idx values to eo-classes.
|
||||
static const unsigned g_sao_eo_idx_to_eo_category[] = { 1, 2, 0, 3, 4 };
|
||||
// Mapping relationships between a, b and c to eo_idx.
|
||||
#define EO_IDX(a, b, c) (2 + SIGN3((c) - (a)) + SIGN3((c) - (b)))
|
||||
|
||||
/**
|
||||
* \param orig_data Original pixel data. 64x64 for luma, 32x32 for chroma.
|
||||
* \param rec_data Reconstructed pixel data. 64x64 for luma, 32x32 for chroma.
|
||||
* \param dir_offsets
|
||||
* \param is_chroma 0 for luma, 1 for chroma. Indicates
|
||||
*/
|
||||
void calc_sao_edge_dir(const pixel *orig_data, const pixel *rec_data,
|
||||
int eo_class, int block_width,
|
||||
int cat_sum_cnt[2][NUM_SAO_EDGE_CATEGORIES])
|
||||
{
|
||||
int y, x;
|
||||
vector2d a_ofs = g_sao_edge_offsets[eo_class][0];
|
||||
vector2d b_ofs = g_sao_edge_offsets[eo_class][1];
|
||||
// Arrays orig_data and rec_data are quarter size for chroma.
|
||||
|
||||
// Don't sample the edge pixels because this function doesn't have access to
|
||||
// their neighbours.
|
||||
for (y = 1; y < block_width - 1; ++y) {
|
||||
for (x = 1; x < block_width - 1; ++x) {
|
||||
const pixel *c_data = &rec_data[y * block_width + x];
|
||||
pixel a = c_data[a_ofs.y * block_width + a_ofs.x];
|
||||
pixel c = c_data[0];
|
||||
pixel b = c_data[b_ofs.y * block_width + b_ofs.x];
|
||||
|
||||
int eo_idx = EO_IDX(a, b, c);
|
||||
int eo_cat = g_sao_eo_idx_to_eo_category[eo_idx];
|
||||
|
||||
cat_sum_cnt[0][eo_cat] += orig_data[y * block_width + x] - c;
|
||||
cat_sum_cnt[1][eo_cat] += 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void sao_reconstruct_color(pixel *rec_data, const sao_info *sao, color_index color)
|
||||
{
|
||||
unsigned y, x;
|
||||
vector2d a_ofs = g_sao_edge_offsets[sao->eo_class][0];
|
||||
vector2d b_ofs = g_sao_edge_offsets[sao->eo_class][1];
|
||||
// Arrays orig_data and rec_data are quarter size for chroma.
|
||||
unsigned block_width = LCU_WIDTH >> !(color == COLOR_Y);
|
||||
|
||||
for (y = 0; y < block_width; ++y) {
|
||||
for (x = 0; x < block_width; ++x) {
|
||||
pixel *c_data = &rec_data[y * block_width + x];
|
||||
pixel a = c_data[a_ofs.y * block_width + a_ofs.x];
|
||||
pixel c = c_data[0];
|
||||
pixel b = c_data[b_ofs.y * block_width + b_ofs.x];
|
||||
|
||||
int eo_idx = EO_IDX(a, b, c);
|
||||
int eo_cat = g_sao_eo_idx_to_eo_category[eo_idx];
|
||||
|
||||
c_data[0] += sao->offsets[eo_cat];
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void sao_reconstruct(picture *pic, unsigned x_ctb, unsigned y_ctb,
|
||||
const sao_info *sao_luma, const sao_info *sao_chroma)
|
||||
{
|
||||
pixel rec_y[LCU_LUMA_SIZE];
|
||||
pixel *y_recdata = &pic->y_recdata[CU_TO_PIXEL(x_ctb, y_ctb, 0, pic->width)];
|
||||
// TODO: sao chroma reconstruct
|
||||
|
||||
// Data to tmp buffer.
|
||||
picture_blit_pixels(y_recdata, rec_y, LCU_WIDTH, LCU_WIDTH, pic->width, LCU_WIDTH);
|
||||
|
||||
sao_reconstruct_color(rec_y, sao_luma, COLOR_Y);
|
||||
//sao_reconstruct_color(rec_u, sao_chroma, COLOR_U);
|
||||
//sao_reconstruct_color(rec_v, sao_chroma, COLOR_V);
|
||||
|
||||
// Copy reconstructed block from tmp buffer to rec image.
|
||||
picture_blit_pixels(rec_y, y_recdata, LCU_WIDTH, LCU_WIDTH, LCU_WIDTH, pic->width);
|
||||
}
|
||||
|
||||
void sao_search_best_mode(const pixel *data, const pixel *recdata,
|
||||
unsigned block_width, unsigned buf_size, unsigned buf_cnt,
|
||||
sao_info *sao_out)
|
||||
{
|
||||
sao_eo_class edge_class;
|
||||
// This array is used to calculate the mean offset used to minimize distortion.
|
||||
int cat_sum_cnt[2][NUM_SAO_EDGE_CATEGORIES];
|
||||
memset(cat_sum_cnt, 0, 2 * NUM_SAO_EDGE_CATEGORIES);
|
||||
|
||||
sao_out->ddistortion = 0;
|
||||
|
||||
for (edge_class = SAO_EO0; edge_class <= SAO_EO3; ++edge_class) {
|
||||
int edge_offset[NUM_SAO_EDGE_CATEGORIES];
|
||||
int sum_ddistortion = 0;
|
||||
sao_eo_cat edge_cat;
|
||||
|
||||
// Call calc_sao_edge_dir once for luma and twice for chroma.
|
||||
while (buf_cnt--) {
|
||||
calc_sao_edge_dir(data, recdata, edge_class, block_width, cat_sum_cnt);
|
||||
data += buf_size;
|
||||
recdata += buf_size;
|
||||
}
|
||||
|
||||
for (edge_cat = SAO_EO_CAT1; edge_cat <= SAO_EO_CAT4; ++edge_cat) {
|
||||
int cat_sum = cat_sum_cnt[0][edge_cat];
|
||||
int cat_cnt = cat_sum_cnt[1][edge_cat];
|
||||
|
||||
// The optimum offset can be calculated by getting the minima of the
|
||||
// fast ddistortion estimation formula. The minima is the mean error
|
||||
// and we round that to the nearest integer.
|
||||
int offset = (cat_sum + (cat_cnt >> 1)) / cat_cnt;
|
||||
edge_offset[edge_cat] = offset;
|
||||
// The ddistortion is amount by which the SSE of data changes. It should
|
||||
// be negative for all categories, if offset was chosen correctly.
|
||||
// ddistortion = N * h^2 - 2 * h * E, where N is the number of samples
|
||||
// and E is the sum of errors.
|
||||
// It basically says that all pixels that are not improved by offset
|
||||
// increase increase SSE by h^2 and all pixels that are improved by
|
||||
// offset decrease SSE by h*E.
|
||||
sum_ddistortion += cat_cnt * offset * offset - 2 * offset * cat_sum;
|
||||
}
|
||||
// SAO is not applied for category 0.
|
||||
edge_offset[SAO_EO_CAT0] = 0;
|
||||
|
||||
// Choose the offset class that offers the least error after offset.
|
||||
if (sum_ddistortion < sao_out->ddistortion) {
|
||||
sao_out->eo_class = edge_class;
|
||||
sao_out->ddistortion = sum_ddistortion;
|
||||
memcpy(sao_out->offsets, edge_offset, NUM_SAO_EDGE_CATEGORIES);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
sao_info sao_search_chroma(const picture *pic, unsigned x_ctb, unsigned y_ctb)
|
||||
{
|
||||
sao_info sao;
|
||||
sao.merge_left_flag = 0;
|
||||
sao.merge_up_flag = 0;
|
||||
sao.type = SAO_TYPE_NONE;
|
||||
return sao;
|
||||
}
|
||||
|
||||
sao_info sao_search_luma(const picture *pic, unsigned x_ctb, unsigned y_ctb)
|
||||
{
|
||||
// These buffers are needed only until we switch to a LCU based data
|
||||
// structure for pixels. Then we can give pointers directly to that structure
|
||||
// without making copies.
|
||||
// It's 2-dimensional because sao_search_best_mode takes arguments as arrays.
|
||||
pixel orig_y[LCU_LUMA_SIZE];
|
||||
pixel rec_y[LCU_LUMA_SIZE];
|
||||
pixel *y_data = &pic->y_data[CU_TO_PIXEL(x_ctb, y_ctb, 0, pic->width)];
|
||||
pixel *y_recdata = &pic->y_recdata[CU_TO_PIXEL(x_ctb, y_ctb, 0, pic->width)];
|
||||
sao_info sao_params;
|
||||
|
||||
// Fill temporary buffers with picture data.
|
||||
picture_blit_pixels(y_data, orig_y, LCU_WIDTH, LCU_WIDTH, pic->width, LCU_WIDTH);
|
||||
picture_blit_pixels(y_recdata, rec_y, LCU_WIDTH, LCU_WIDTH, pic->width, LCU_WIDTH);
|
||||
|
||||
sao_search_best_mode(orig_y, rec_y, LCU_WIDTH, LCU_LUMA_SIZE, 1, &sao_params);
|
||||
|
||||
return sao_params;
|
||||
}
|
||||
|
||||
void encode_sao_offsets(encoder_control *encoder, sao_info *sao)
|
||||
{
|
||||
int i;
|
||||
|
|
@ -859,34 +1040,20 @@ void encode_sao_merge_flags(encoder_control *encoder, sao_info *sao,
|
|||
/**
|
||||
* \brief Stub that encodes all LCU's as none type.
|
||||
*/
|
||||
void encode_sao(encoder_control *encoder, unsigned x_lcu, uint16_t y_lcu)
|
||||
void encode_sao(encoder_control *encoder, unsigned x_lcu, uint16_t y_lcu,
|
||||
sao_info *sao_luma, sao_info *sao_chroma)
|
||||
{
|
||||
unsigned sao_type[3] = {SAO_TYPE_NONE, SAO_TYPE_NONE, SAO_TYPE_NONE};
|
||||
picture *pic = encoder->in.cur_pic;
|
||||
sao_info tmp_sao[3];
|
||||
sao_info *sao = &tmp_sao[0];
|
||||
|
||||
// The tmp_sao and these assignments are temporary. The sao pointer will
|
||||
// be given to this function.
|
||||
sao[0].merge_left_flag = 0;
|
||||
sao[0].merge_up_flag = 0;
|
||||
sao[0].type = SAO_TYPE_NONE;
|
||||
|
||||
sao[1].merge_left_flag = 0;
|
||||
sao[1].merge_up_flag = 0;
|
||||
sao[1].type = SAO_TYPE_NONE;
|
||||
|
||||
sao[2].merge_left_flag = 0;
|
||||
sao[2].merge_up_flag = 0;
|
||||
sao[2].type = SAO_TYPE_NONE;
|
||||
|
||||
encode_sao_merge_flags(encoder, sao, x_lcu, y_lcu);
|
||||
// TODO: transmit merge flags outside sao_info
|
||||
encode_sao_merge_flags(encoder, sao_luma, x_lcu, y_lcu);
|
||||
|
||||
// If SAO is merged, nothing else needs to be coded.
|
||||
if (!sao->merge_left_flag && !sao->merge_up_flag) {
|
||||
encode_sao_color(encoder, &sao[COLOR_Y], COLOR_Y);
|
||||
encode_sao_color(encoder, &sao[COLOR_U], COLOR_U);
|
||||
encode_sao_color(encoder, &sao[COLOR_V], COLOR_V);
|
||||
if (!sao_luma->merge_left_flag && !sao_luma->merge_up_flag) {
|
||||
encode_sao_color(encoder, sao_luma, COLOR_Y);
|
||||
encode_sao_color(encoder, sao_chroma, COLOR_U);
|
||||
encode_sao_color(encoder, sao_chroma, COLOR_V);
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -906,7 +1073,12 @@ void encode_slice_data(encoder_control* encoder)
|
|||
uint8_t depth = 0;
|
||||
|
||||
if (encoder->sao_enable) {
|
||||
encode_sao(encoder, x_ctb, y_ctb);
|
||||
sao_info sao_luma = sao_search_luma(encoder->in.cur_pic, x_ctb, y_ctb);
|
||||
sao_info sao_chroma = sao_search_chroma(encoder->in.cur_pic, x_ctb, y_ctb);
|
||||
|
||||
// sao_do_merge(encoder, x_ctb, y_ctb, sao_luma, sao_chroma);
|
||||
// sao_do_rdo(encoder, x_ctb, y_ctb, sao_luma, sao_chroma);
|
||||
encode_sao(encoder, x_ctb, y_ctb, &sao_luma, &sao_chroma);
|
||||
}
|
||||
|
||||
// Recursive function for looping through all the sub-blocks
|
||||
|
|
|
|||
|
|
@ -65,6 +65,9 @@ typedef int16_t coefficient;
|
|||
|
||||
/* END OF CONFIG VARIABLES */
|
||||
|
||||
#define LCU_LUMA_SIZE (LCU_WIDTH * LCU_WIDTH)
|
||||
#define LCU_CHROMA_SIZE (LCU_WIDTH * LCU_WIDTH >> 2)
|
||||
|
||||
#define MAX_REF_PIC_COUNT 5
|
||||
|
||||
#define AMVP_MAX_NUM_CANDS 2
|
||||
|
|
@ -80,6 +83,12 @@ typedef int16_t coefficient;
|
|||
#define NO_SCU_IN_LCU(no_lcu) ((no_lcu) << MAX_DEPTH)
|
||||
#define WITHIN(val, min_val, max_val) ((min_val) <= (val) && (val) <= (max_val))
|
||||
|
||||
#define LOG2_LCU_WIDTH 6
|
||||
// CU_TO_PIXEL = y * lcu_width * pic_width + x * lcu_width
|
||||
#define CU_TO_PIXEL(x, y, depth, width) (((y) << (LOG2_LCU_WIDTH - (depth))) * (width) \
|
||||
+ ((x) << (LOG2_LCU_WIDTH - (depth))))
|
||||
|
||||
|
||||
#define VERSION_STRING "0.2 "
|
||||
#define VERSION 0.2
|
||||
|
||||
|
|
|
|||
|
|
@ -43,6 +43,37 @@ void picture_set_block_residual(picture *pic, uint32_t x_scu, uint32_t y_scu,
|
|||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* \brief BLock Image Transfer from one buffer to another.
|
||||
*
|
||||
* It's a stupidly simple loop that copies pixels.
|
||||
*
|
||||
* \param orig Start of the originating buffer.
|
||||
* \param dst Start of the destination buffer.
|
||||
* \param width Width of the copied region.
|
||||
* \param height Height of the copied region.
|
||||
* \param orig_stride Width of a row in the originating buffer.
|
||||
* \param dst_stride Width of a row in the destination buffer.
|
||||
*
|
||||
* This should be inlined, but it's defined here for now to see if Visual
|
||||
* Studios LTCG will inline it.
|
||||
*/
|
||||
void picture_blit_pixels(const pixel* orig, pixel *dst,
|
||||
unsigned width, unsigned height,
|
||||
unsigned orig_stride, unsigned dst_stride)
|
||||
{
|
||||
unsigned y, x;
|
||||
|
||||
for (y = 0; y < height; ++y) {
|
||||
for (x = 0; x < width; ++x) {
|
||||
dst[x] = orig[x];
|
||||
}
|
||||
// Move pointers to the next row.
|
||||
orig += orig_stride;
|
||||
dst += dst_stride;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* \brief Set block coded status
|
||||
* \param pic picture to use
|
||||
|
|
|
|||
|
|
@ -108,6 +108,9 @@ void picture_set_block_residual(picture *pic, uint32_t x_scu, uint32_t y_scu,
|
|||
uint8_t depth, int8_t residual);
|
||||
void picture_set_block_split(picture *pic, uint32_t x_scu, uint32_t y_scu,
|
||||
uint8_t depth, int8_t split);
|
||||
void picture_blit_pixels(const pixel* orig, pixel *dst,
|
||||
unsigned width, unsigned height,
|
||||
unsigned orig_stride, unsigned dst_stride);
|
||||
|
||||
picture_list * picture_list_init(int size);
|
||||
int picture_list_resize(picture_list *list, int size);
|
||||
|
|
|
|||
Loading…
Reference in a new issue