Fix frame scaling prediction
Use higher precision offsets for more accurate predictor
generation when references are at a different scale from
the coded frame.
Change-Id: I4c2c0ec67fa4824273cb3bd072211f41ac7802e8
diff --git a/aom_dsp/aom_convolve.c b/aom_dsp/aom_convolve.c
index 4dac6aa..52bb542 100644
--- a/aom_dsp/aom_convolve.c
+++ b/aom_dsp/aom_convolve.c
@@ -41,6 +41,29 @@
}
}
+static void convolve_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *x_filters, int x0_qn,
+ int x_step_qn, int w, int h) {
+ int x, y;
+ src -= SUBPEL_TAPS / 2 - 1;
+ for (y = 0; y < h; ++y) {
+ int x_qn = x0_qn;
+ for (x = 0; x < w; ++x) {
+ const uint8_t *const src_x = &src[x_qn >> SCALE_SUBPEL_BITS]; // q8
+ const int x_filter_idx = (x_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS;
+ assert(x_filter_idx < SUBPEL_SHIFTS);
+ const int16_t *const x_filter = x_filters[x_filter_idx];
+ int k, sum = 0;
+ for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
+ dst[x] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
+ x_qn += x_step_qn;
+ }
+ src += src_stride;
+ dst += dst_stride;
+ }
+}
+
static void convolve_avg_horiz(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const InterpKernel *x_filters, int x0_q4,
@@ -63,6 +86,30 @@
}
}
+static void convolve_avg_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *x_filters, int x0_qn,
+ int x_step_qn, int w, int h) {
+ int x, y;
+ src -= SUBPEL_TAPS / 2 - 1;
+ for (y = 0; y < h; ++y) {
+ int x_qn = x0_qn;
+ for (x = 0; x < w; ++x) {
+ const uint8_t *const src_x = &src[x_qn >> SCALE_SUBPEL_BITS];
+ const int x_filter_idx = (x_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS;
+ assert(x_filter_idx < SUBPEL_SHIFTS);
+ const int16_t *const x_filter = x_filters[x_filter_idx];
+ int k, sum = 0;
+ for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
+ dst[x] = ROUND_POWER_OF_TWO(
+ dst[x] + clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)), 1);
+ x_qn += x_step_qn;
+ }
+ src += src_stride;
+ dst += dst_stride;
+ }
+}
+
static void convolve_vert(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const InterpKernel *y_filters, int y0_q4,
@@ -86,6 +133,31 @@
}
}
+static void convolve_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *y_filters, int y0_qn,
+ int y_step_qn, int w, int h) {
+ int x, y;
+ src -= src_stride * (SUBPEL_TAPS / 2 - 1);
+
+ for (x = 0; x < w; ++x) {
+ int y_qn = y0_qn;
+ for (y = 0; y < h; ++y) {
+ const unsigned char *src_y =
+ &src[(y_qn >> SCALE_SUBPEL_BITS) * src_stride];
+ const int16_t *const y_filter =
+ y_filters[(y_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS];
+ int k, sum = 0;
+ for (k = 0; k < SUBPEL_TAPS; ++k)
+ sum += src_y[k * src_stride] * y_filter[k];
+ dst[y * dst_stride] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
+ y_qn += y_step_qn;
+ }
+ ++src;
+ ++dst;
+ }
+}
+
static void convolve_avg_vert(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const InterpKernel *y_filters, int y0_q4,
@@ -112,6 +184,34 @@
}
}
+static void convolve_avg_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *y_filters, int y0_qn,
+ int y_step_qn, int w, int h) {
+ int x, y;
+ src -= src_stride * (SUBPEL_TAPS / 2 - 1);
+
+ for (x = 0; x < w; ++x) {
+ int y_qn = y0_qn;
+ for (y = 0; y < h; ++y) {
+ const unsigned char *src_y =
+ &src[(y_qn >> SCALE_SUBPEL_BITS) * src_stride];
+ const int16_t *const y_filter =
+ y_filters[(y_qn & SCALE_SUBPEL_MASK) >> SCALE_EXTRA_BITS];
+ int k, sum = 0;
+ for (k = 0; k < SUBPEL_TAPS; ++k)
+ sum += src_y[k * src_stride] * y_filter[k];
+ dst[y * dst_stride] = ROUND_POWER_OF_TWO(
+ dst[y * dst_stride] +
+ clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)),
+ 1);
+ y_qn += y_step_qn;
+ }
+ ++src;
+ ++dst;
+ }
+}
+
static void convolve(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
ptrdiff_t dst_stride, const InterpKernel *const x_filters,
int x0_q4, int x_step_q4,
@@ -146,6 +246,41 @@
dst_stride, y_filters, y0_q4, y_step_q4, w, h);
}
+static void convolve_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const InterpKernel *const x_filters, int x0_qn,
+ int x_step_qn, const InterpKernel *const y_filters,
+ int y0_qn, int y_step_qn, int w, int h) {
+ // TODO(afergs): Update comment here
+ // Note: Fixed size intermediate buffer, temp, places limits on parameters.
+ // 2d filtering proceeds in 2 steps:
+ // (1) Interpolate horizontally into an intermediate buffer, temp.
+ // (2) Interpolate temp vertically to derive the sub-pixel result.
+ // Deriving the maximum number of rows in the temp buffer (135):
+ // --Smallest scaling factor is x1/2 ==> y_step_qn = 32 (Normative).
+ // --Largest block size is 64x64 pixels.
+ // --64 rows in the downscaled frame span a distance of (64 - 1) * 32 in the
+ // original frame (in 1/16th pixel units).
+ // --Must round-up because block may be located at sub-pixel position.
+ // --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
+ // --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
+ uint8_t temp[MAX_EXT_SIZE * MAX_SB_SIZE];
+ int intermediate_height =
+ (((h - 1) * y_step_qn + y0_qn) >> SCALE_SUBPEL_BITS) + SUBPEL_TAPS;
+
+ assert(w <= MAX_SB_SIZE);
+ assert(h <= MAX_SB_SIZE);
+
+ assert(y_step_qn <= SCALE_SUBPEL_BITS * 2);
+ assert(x_step_qn <= SCALE_SUBPEL_BITS * 2);
+
+ convolve_horiz_scale_c(src - src_stride * (SUBPEL_TAPS / 2 - 1), src_stride,
+ temp, MAX_SB_SIZE, x_filters, x0_qn, x_step_qn, w,
+ intermediate_height);
+ convolve_vert_scale_c(temp + MAX_SB_SIZE * (SUBPEL_TAPS / 2 - 1), MAX_SB_SIZE,
+ dst, dst_stride, y_filters, y0_qn, y_step_qn, w, h);
+}
+
static const InterpKernel *get_filter_base(const int16_t *filter) {
// NOTE: This assumes that the filter table is 256-byte aligned.
// TODO(agrange) Modify to make independent of table alignment.
@@ -171,6 +306,21 @@
w, h);
}
+void aom_convolve8_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x,
+ int x_step_qn, const int16_t *filter_y,
+ int subpel_y, int y_step_qn, int w, int h) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+
+ (void)subpel_y;
+ (void)filter_y;
+ (void)y_step_qn;
+
+ convolve_horiz_scale_c(src, src_stride, dst, dst_stride, filters_x, subpel_x,
+ x_step_qn, w, h);
+}
+
void aom_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const int16_t *filter_x, int x_step_q4,
@@ -186,6 +336,22 @@
x_step_q4, w, h);
}
+void aom_convolve8_avg_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x,
+ int x_step_qn, const int16_t *filter_y,
+ int subpel_y, int y_step_qn, int w,
+ int h) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+
+ (void)subpel_y;
+ (void)filter_y;
+ (void)y_step_qn;
+
+ convolve_avg_horiz_scale_c(src, src_stride, dst, dst_stride, filters_x,
+ subpel_x, x_step_qn, w, h);
+}
+
void aom_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const int16_t *filter_x, int x_step_q4,
@@ -201,6 +367,21 @@
w, h);
}
+void aom_convolve8_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x,
+ int x_step_qn, const int16_t *filter_y,
+ int subpel_y, int y_step_qn, int w, int h) {
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+
+ (void)subpel_x;
+ (void)filter_x;
+ (void)x_step_qn;
+
+ convolve_vert_scale_c(src, src_stride, dst, dst_stride, filters_y, subpel_y,
+ y_step_qn, w, h);
+}
+
void aom_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
const int16_t *filter_x, int x_step_q4,
@@ -216,6 +397,21 @@
y_step_q4, w, h);
}
+void aom_convolve8_avg_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x,
+ int x_step_qn, const int16_t *filter_y,
+ int subpel_y, int y_step_qn, int w, int h) {
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+
+ (void)subpel_x;
+ (void)filter_x;
+ (void)x_step_qn;
+
+ convolve_avg_vert_scale_c(src, src_stride, dst, dst_stride, filters_y,
+ subpel_y, y_step_qn, w, h);
+}
+
void aom_convolve8_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
ptrdiff_t dst_stride, const int16_t *filter_x,
int x_step_q4, const int16_t *filter_y, int y_step_q4,
@@ -230,6 +426,19 @@
filters_y, y0_q4, y_step_q4, w, h);
}
+void aom_convolve8_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x, int x_step_qn,
+ const int16_t *filter_y, int subpel_y, int y_step_qn,
+ int w, int h) {
+ const InterpKernel *const filters_x = get_filter_base(filter_x);
+
+ const InterpKernel *const filters_y = get_filter_base(filter_y);
+
+ convolve_scale_c(src, src_stride, dst, dst_stride, filters_x, subpel_x,
+ x_step_qn, filters_y, subpel_y, y_step_qn, w, h);
+}
+
void aom_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
ptrdiff_t dst_stride, const int16_t *filter_x,
int x_step_q4, const int16_t *filter_y, int y_step_q4,
@@ -245,6 +454,22 @@
h);
}
+void aom_convolve8_avg_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int subpel_x,
+ int x_step_qn, const int16_t *filter_y,
+ int subpel_y, int y_step_qn, int w, int h) {
+ /* Fixed size intermediate buffer places limits on parameters. */
+ DECLARE_ALIGNED(16, uint8_t, temp[MAX_SB_SIZE * MAX_SB_SIZE]);
+ assert(w <= MAX_SB_SIZE);
+ assert(h <= MAX_SB_SIZE);
+
+ aom_convolve8_scale_c(src, src_stride, temp, MAX_SB_SIZE, filter_x, subpel_x,
+ x_step_qn, filter_y, subpel_y, y_step_qn, w, h);
+ aom_convolve_avg_c(temp, MAX_SB_SIZE, dst, dst_stride, NULL, 0, NULL, 0, w,
+ h);
+}
+
void aom_convolve_copy_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
ptrdiff_t dst_stride, const int16_t *filter_x,
int filter_x_stride, const int16_t *filter_y,
@@ -332,6 +557,63 @@
filter_y, y_step_q4, w, h);
}
+/*
+void aom_scaled_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_horiz_scale_c(src, src_stride, dst, dst_stride, filter_x,
+ x_step_qn, filter_y, y_step_qn, w, h);
+}
+
+void aom_scaled_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_vert_scale_c(src, src_stride, dst, dst_stride, filter_x,
+ x_step_qn, filter_y, y_step_qn, w, h);
+}
+
+void aom_scaled_2d_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_scale_c(src, src_stride, dst, dst_stride, filter_x, x_step_qn,
+ filter_y, y_step_qn, w, h);
+}
+
+void aom_scaled_avg_horiz_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_avg_horiz_scale_c(src, src_stride, dst, dst_stride, filter_x,
+ x_step_qn, filter_y, y_step_qn, w, h);
+}
+
+void aom_scaled_avg_vert_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_avg_vert_scale_c(src, src_stride, dst, dst_stride, filter_x,
+ x_step_qn, filter_y, y_step_qn, w, h);
+}
+
+void aom_scaled_avg_2d_scale_c(const uint8_t *src, ptrdiff_t src_stride,
+ uint8_t *dst, ptrdiff_t dst_stride,
+ const int16_t *filter_x, int x_step_qn,
+ const int16_t *filter_y, int y_step_qn, int w,
+ int h) {
+ aom_convolve8_avg_scale_c(src, src_stride, dst, dst_stride, filter_x,
+ x_step_qn, filter_y, y_step_qn, w, h);
+}
+*/
+
+// TODO(afergs): Make sure this works too
#if CONFIG_LOOP_RESTORATION
static void convolve_add_src_horiz(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
@@ -569,6 +851,7 @@
}
#endif // CONFIG_LOOP_RESTORATION
+// TODO(afergs): Make sure this works too
#if CONFIG_HIGHBITDEPTH
static void highbd_convolve_horiz(const uint8_t *src8, ptrdiff_t src_stride,
uint8_t *dst8, ptrdiff_t dst_stride,
