mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-03 16:09:39 +00:00
631 lines
15 KiB
C
631 lines
15 KiB
C
|
/*
|
||
|
* Image Scaling Functions
|
||
|
* Copyright (c) 2005 David A. Schleef <ds@schleef.org>
|
||
|
* All rights reserved.
|
||
|
*
|
||
|
* Redistribution and use in source and binary forms, with or without
|
||
|
* modification, are permitted provided that the following conditions
|
||
|
* are met:
|
||
|
* 1. Redistributions of source code must retain the above copyright
|
||
|
* notice, this list of conditions and the following disclaimer.
|
||
|
* 2. Redistributions in binary form must reproduce the above copyright
|
||
|
* notice, this list of conditions and the following disclaimer in the
|
||
|
* documentation and/or other materials provided with the distribution.
|
||
|
*
|
||
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
|
||
|
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
||
|
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
||
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
|
||
|
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
|
||
|
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
||
|
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
||
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
||
|
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
|
||
|
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
||
|
* POSSIBILITY OF SUCH DAMAGE.
|
||
|
*/
|
||
|
|
||
|
#include "vs_scanline.h"
|
||
|
|
||
|
#include <glib.h>
|
||
|
|
||
|
/* greyscale, i.e., single componenet */
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_Y (guint8 * dest, guint8 * src, int n)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = (src[i * 2] + src[i * 2 + 1]) / 2;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_Y (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = (x < 32768) ? src[j] : src[j + 1];
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_Y (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = (src[j] * (65536 - x) + src[j + 1] * x) >> 16;
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_Y (guint8 * dest, guint8 * src1, guint8 * src2,
|
||
|
int n, int x)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = (src1[i] * (65536 - x) + src2[i] * x) >> 16;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* RGBA */
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_RGBA (guint8 * dest, guint8 * src, int n)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] = (src[i * 8 + 0] + src[i * 8 + 4]) / 2;
|
||
|
dest[i * 4 + 1] = (src[i * 8 + 1] + src[i * 8 + 5]) / 2;
|
||
|
dest[i * 4 + 2] = (src[i * 8 + 2] + src[i * 8 + 6]) / 2;
|
||
|
dest[i * 4 + 3] = (src[i * 8 + 3] + src[i * 8 + 7]) / 2;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_RGBA (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 0] = (x < 32768) ? src[j * 4 + 0] : src[j * 4 + 4];
|
||
|
dest[i * 4 + 1] = (x < 32768) ? src[j * 4 + 1] : src[j * 4 + 5];
|
||
|
dest[i * 4 + 2] = (x < 32768) ? src[j * 4 + 2] : src[j * 4 + 6];
|
||
|
dest[i * 4 + 3] = (x < 32768) ? src[j * 4 + 3] : src[j * 4 + 7];
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_RGBA (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 0] = (src[j * 4 + 0] * (65536 - x) + src[j * 4 + 4] * x) >> 16;
|
||
|
dest[i * 4 + 1] = (src[j * 4 + 1] * (65536 - x) + src[j * 4 + 5] * x) >> 16;
|
||
|
dest[i * 4 + 2] = (src[j * 4 + 2] * (65536 - x) + src[j * 4 + 6] * x) >> 16;
|
||
|
dest[i * 4 + 3] = (src[j * 4 + 3] * (65536 - x) + src[j * 4 + 7] * x) >> 16;
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_RGBA (guint8 * dest, guint8 * src1, guint8 * src2,
|
||
|
int n, int x)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] =
|
||
|
(src1[i * 4 + 0] * (65536 - x) + src2[i * 4 + 0] * x) >> 16;
|
||
|
dest[i * 4 + 1] =
|
||
|
(src1[i * 4 + 1] * (65536 - x) + src2[i * 4 + 1] * x) >> 16;
|
||
|
dest[i * 4 + 2] =
|
||
|
(src1[i * 4 + 2] * (65536 - x) + src2[i * 4 + 2] * x) >> 16;
|
||
|
dest[i * 4 + 3] =
|
||
|
(src1[i * 4 + 3] * (65536 - x) + src2[i * 4 + 3] * x) >> 16;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* RGB */
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_RGB (guint8 * dest, guint8 * src, int n)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 3 + 0] = (src[i * 6 + 0] + src[i * 6 + 3]) / 2;
|
||
|
dest[i * 3 + 1] = (src[i * 6 + 1] + src[i * 6 + 4]) / 2;
|
||
|
dest[i * 3 + 2] = (src[i * 6 + 2] + src[i * 6 + 5]) / 2;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_RGB (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 3 + 0] = (x < 32768) ? src[j * 3 + 0] : src[j * 3 + 3];
|
||
|
dest[i * 3 + 1] = (x < 32768) ? src[j * 3 + 1] : src[j * 3 + 4];
|
||
|
dest[i * 3 + 2] = (x < 32768) ? src[j * 3 + 2] : src[j * 3 + 5];
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_RGB (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 3 + 0] = (src[j * 3 + 0] * (65536 - x) + src[j * 3 + 3] * x) >> 16;
|
||
|
dest[i * 3 + 1] = (src[j * 3 + 1] * (65536 - x) + src[j * 3 + 4] * x) >> 16;
|
||
|
dest[i * 3 + 2] = (src[j * 3 + 2] * (65536 - x) + src[j * 3 + 5] * x) >> 16;
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_RGB (guint8 * dest, guint8 * src1, guint8 * src2,
|
||
|
int n, int x)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 3 + 0] =
|
||
|
(src1[i * 3 + 0] * (65536 - x) + src2[i * 3 + 3] * x) >> 16;
|
||
|
dest[i * 3 + 1] =
|
||
|
(src1[i * 3 + 1] * (65536 - x) + src2[i * 3 + 4] * x) >> 16;
|
||
|
dest[i * 3 + 2] =
|
||
|
(src1[i * 3 + 2] * (65536 - x) + src2[i * 3 + 5] * x) >> 16;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* YUYV */
|
||
|
|
||
|
/* n is the number of bi-pixels */
|
||
|
/* increment is per Y pixel */
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_YUYV (guint8 * dest, guint8 * src, int n)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] = (src[i * 8 + 0] + src[i * 8 + 2]) / 2;
|
||
|
dest[i * 4 + 1] = (src[i * 8 + 1] + src[i * 8 + 5]) / 2;
|
||
|
dest[i * 4 + 2] = (src[i * 8 + 4] + src[i * 8 + 6]) / 2;
|
||
|
dest[i * 4 + 3] = (src[i * 8 + 3] + src[i * 8 + 7]) / 2;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_YUYV (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 0] = (x < 32768) ? src[j * 2 + 0] : src[j * 2 + 2];
|
||
|
|
||
|
j = acc >> 17;
|
||
|
x = acc & 0x1ffff;
|
||
|
dest[i * 4 + 1] = (x < 65536) ? src[j * 4 + 1] : src[j * 4 + 5];
|
||
|
dest[i * 4 + 3] = (x < 65536) ? src[j * 4 + 3] : src[j * 4 + 7];
|
||
|
|
||
|
acc += increment;
|
||
|
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 2] = (x < 32768) ? src[j * 2 + 0] : src[j * 2 + 2];
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_YUYV (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 0] = (src[j * 2 + 0] * (65536 - x) + src[j * 2 + 2] * x) >> 16;
|
||
|
|
||
|
j = acc >> 17;
|
||
|
x = acc & 0x1ffff;
|
||
|
dest[i * 4 + 1] =
|
||
|
(src[j * 4 + 1] * (131072 - x) + src[j * 4 + 5] * x) >> 17;
|
||
|
dest[i * 4 + 3] =
|
||
|
(src[j * 4 + 3] * (131072 - x) + src[j * 4 + 7] * x) >> 17;
|
||
|
|
||
|
acc += increment;
|
||
|
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 2] = (src[j * 2 + 0] * (65536 - x) + src[j * 2 + 2] * x) >> 16;
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_YUYV (guint8 * dest, guint8 * src1, guint8 * src2,
|
||
|
int n, int x)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] =
|
||
|
(src1[i * 4 + 0] * (65536 - x) + src2[i * 4 + 0] * x) >> 16;
|
||
|
dest[i * 4 + 1] =
|
||
|
(src1[i * 4 + 1] * (65536 - x) + src2[i * 4 + 1] * x) >> 16;
|
||
|
dest[i * 4 + 2] =
|
||
|
(src1[i * 4 + 2] * (65536 - x) + src2[i * 4 + 2] * x) >> 16;
|
||
|
dest[i * 4 + 3] =
|
||
|
(src1[i * 4 + 3] * (65536 - x) + src2[i * 4 + 3] * x) >> 16;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* UYVY */
|
||
|
|
||
|
/* n is the number of bi-pixels */
|
||
|
/* increment is per Y pixel */
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_UYVY (guint8 * dest, guint8 * src, int n)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] = (src[i * 8 + 0] + src[i * 8 + 4]) / 2;
|
||
|
dest[i * 4 + 1] = (src[i * 8 + 1] + src[i * 8 + 3]) / 2;
|
||
|
dest[i * 4 + 2] = (src[i * 8 + 2] + src[i * 8 + 6]) / 2;
|
||
|
dest[i * 4 + 3] = (src[i * 8 + 5] + src[i * 8 + 7]) / 2;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_UYVY (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 1] = (x < 32768) ? src[j * 2 + 1] : src[j * 2 + 3];
|
||
|
|
||
|
j = acc >> 17;
|
||
|
x = acc & 0x1ffff;
|
||
|
dest[i * 4 + 0] = (x < 65536) ? src[j * 4 + 0] : src[j * 4 + 4];
|
||
|
dest[i * 4 + 2] = (x < 65536) ? src[j * 4 + 2] : src[j * 4 + 6];
|
||
|
|
||
|
acc += increment;
|
||
|
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 3] = (x < 32768) ? src[j * 2 + 1] : src[j * 2 + 3];
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_UYVY (guint8 * dest, guint8 * src, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 1] = (src[j * 2 + 1] * (65536 - x) + src[j * 2 + 3] * x) >> 16;
|
||
|
|
||
|
j = acc >> 17;
|
||
|
x = acc & 0x1ffff;
|
||
|
dest[i * 4 + 0] =
|
||
|
(src[j * 4 + 0] * (131072 - x) + src[j * 4 + 4] * x) >> 17;
|
||
|
dest[i * 4 + 2] =
|
||
|
(src[j * 4 + 2] * (131072 - x) + src[j * 4 + 6] * x) >> 17;
|
||
|
|
||
|
acc += increment;
|
||
|
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i * 4 + 3] = (src[j * 2 + 1] * (65536 - x) + src[j * 2 + 3] * x) >> 16;
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_UYVY (guint8 * dest, guint8 * src1, guint8 * src2,
|
||
|
int n, int x)
|
||
|
{
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i * 4 + 0] =
|
||
|
(src1[i * 4 + 0] * (65536 - x) + src2[i * 4 + 0] * x) >> 16;
|
||
|
dest[i * 4 + 1] =
|
||
|
(src1[i * 4 + 1] * (65536 - x) + src2[i * 4 + 1] * x) >> 16;
|
||
|
dest[i * 4 + 2] =
|
||
|
(src1[i * 4 + 2] * (65536 - x) + src2[i * 4 + 2] * x) >> 16;
|
||
|
dest[i * 4 + 3] =
|
||
|
(src1[i * 4 + 3] * (65536 - x) + src2[i * 4 + 3] * x) >> 16;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* RGB565 */
|
||
|
|
||
|
/* note that src and dest are guint16, and thus endian dependent */
|
||
|
|
||
|
#define RGB565_R(x) (((x)&0xf800)>>8 | ((x)&0xf800)>>13)
|
||
|
#define RGB565_G(x) (((x)&0x07e0)>>3 | ((x)&0x07e0)>>9)
|
||
|
#define RGB565_B(x) (((x)&0x001f)<<3 | ((x)&0x001f)>>2)
|
||
|
|
||
|
#define RGB565(r,g,b) \
|
||
|
((((r)<<8)&0xf800) | (((g)<<3)&0x07e0) | (((b)>>3)&0x001f))
|
||
|
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_RGB565 (guint8 * dest_u8, guint8 * src_u8, int n)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = RGB565 (
|
||
|
(RGB565_R (src[i * 2]) + RGB565_R (src[i * 2 + 1])) / 2,
|
||
|
(RGB565_G (src[i * 2]) + RGB565_G (src[i * 2 + 1])) / 2,
|
||
|
(RGB565_B (src[i * 2]) + RGB565_B (src[i * 2 + 1])) / 2);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_RGB565 (guint8 * dest_u8, guint8 * src_u8, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = (x < 32768) ? src[j] : src[j + 1];
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_RGB565 (guint8 * dest_u8, guint8 * src_u8, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = RGB565 (
|
||
|
(RGB565_R (src[j]) * (65536 - x) + RGB565_R (src[j + 1]) * x) >> 16,
|
||
|
(RGB565_G (src[j]) * (65536 - x) + RGB565_G (src[j + 1]) * x) >> 16,
|
||
|
(RGB565_B (src[j]) * (65536 - x) + RGB565_B (src[j + 1]) * x) >> 16);
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_RGB565 (guint8 * dest_u8, guint8 * src1_u8,
|
||
|
guint8 * src2_u8, int n, int x)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src1 = (guint16 *) src1_u8;
|
||
|
guint16 *src2 = (guint16 *) src2_u8;
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = RGB565 (
|
||
|
(RGB565_R (src1[i]) * (65536 - x) + RGB565_R (src2[i]) * x) >> 16,
|
||
|
(RGB565_G (src1[i]) * (65536 - x) + RGB565_G (src2[i]) * x) >> 16,
|
||
|
(RGB565_B (src1[i]) * (65536 - x) + RGB565_B (src2[i]) * x) >> 16);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
|
||
|
/* RGB555 */
|
||
|
|
||
|
/* note that src and dest are guint16, and thus endian dependent */
|
||
|
|
||
|
#define RGB555_R(x) (((x)&0x7c00)>>8 | ((x)&0x7c00)>>13)
|
||
|
#define RGB555_G(x) (((x)&0x03e0)>>3 | ((x)&0x03e0)>>9)
|
||
|
#define RGB555_B(x) (((x)&0x001f)<<3 | ((x)&0x001f)>>2)
|
||
|
|
||
|
#define RGB555(r,g,b) \
|
||
|
((((r)<<7)&0x7c00) | (((g)<<3)&0x03e0) | (((b)>>3)&0x001f))
|
||
|
|
||
|
|
||
|
void
|
||
|
vs_scanline_downsample_RGB555 (guint8 * dest_u8, guint8 * src_u8, int n)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = RGB555 (
|
||
|
(RGB555_R (src[i * 2]) + RGB555_R (src[i * 2 + 1])) / 2,
|
||
|
(RGB555_G (src[i * 2]) + RGB555_G (src[i * 2 + 1])) / 2,
|
||
|
(RGB555_B (src[i * 2]) + RGB555_B (src[i * 2 + 1])) / 2);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_nearest_RGB555 (guint8 * dest_u8, guint8 * src_u8, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = (x < 32768) ? src[j] : src[j + 1];
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_resample_linear_RGB555 (guint8 * dest_u8, guint8 * src_u8, int n,
|
||
|
int *accumulator, int increment)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src = (guint16 *) src_u8;
|
||
|
int acc = *accumulator;
|
||
|
int i;
|
||
|
int j;
|
||
|
int x;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
j = acc >> 16;
|
||
|
x = acc & 0xffff;
|
||
|
dest[i] = RGB555 (
|
||
|
(RGB555_R (src[j]) * (65536 - x) + RGB555_R (src[j + 1]) * x) >> 16,
|
||
|
(RGB555_G (src[j]) * (65536 - x) + RGB555_G (src[j + 1]) * x) >> 16,
|
||
|
(RGB555_B (src[j]) * (65536 - x) + RGB555_B (src[j + 1]) * x) >> 16);
|
||
|
|
||
|
acc += increment;
|
||
|
}
|
||
|
|
||
|
*accumulator = acc;
|
||
|
}
|
||
|
|
||
|
void
|
||
|
vs_scanline_merge_linear_RGB555 (guint8 * dest_u8, guint8 * src1_u8,
|
||
|
guint8 * src2_u8, int n, int x)
|
||
|
{
|
||
|
guint16 *dest = (guint16 *) dest_u8;
|
||
|
guint16 *src1 = (guint16 *) src1_u8;
|
||
|
guint16 *src2 = (guint16 *) src2_u8;
|
||
|
int i;
|
||
|
|
||
|
for (i = 0; i < n; i++) {
|
||
|
dest[i] = RGB555 (
|
||
|
(RGB555_R (src1[i]) * (65536 - x) + RGB555_R (src2[i]) * x) >> 16,
|
||
|
(RGB555_G (src1[i]) * (65536 - x) + RGB555_G (src2[i]) * x) >> 16,
|
||
|
(RGB555_B (src1[i]) * (65536 - x) + RGB555_B (src2[i]) * x) >> 16);
|
||
|
}
|
||
|
}
|