Synchronet now requires the libarchive development package (e.g. libarchive-dev on Debian-based Linux distros, libarchive.org for more info) to build successfully.

Commits (1)
  • Deucе's avatar
    Add smooth scaling to X11 output · cae45cbd
    Deucе authored
    Uses "pointyscale" for x3 and x5
    Uses xBR from FFmpeg for x2 and x4
    HQx is also included, but unused as it's too slow at x4.
    cae45cbd
......@@ -15,7 +15,9 @@ ifdef NO_X
CFLAGS += -DNO_X
else
OBJS += $(MTOBJODIR)$(DIRSEP)x_events$(OFILE) \
$(MTOBJODIR)$(DIRSEP)x_cio$(OFILE)
$(MTOBJODIR)$(DIRSEP)x_cio$(OFILE) \
$(MTOBJODIR)$(DIRSEP)scale$(OFILE) \
$(MTOBJODIR)$(DIRSEP)xbr$(OFILE)
NEED_BITMAP := 1
endif
......
/*
* Copyright (c) 2014 Clément Bœsch
*
* This file is part of FFmpeg.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
/**
* @file
* hqx magnification filters (hq2x, hq3x, hq4x)
*
* Originally designed by Maxim Stephin.
*
* @see http://en.wikipedia.org/wiki/Hqx
* @see http://web.archive.org/web/20131114143602/http://www.hiend3d.com/hq3x.html
* @see http://blog.pkh.me/p/19-butchering-hqx-scaling-filters.html
*/
#include <inttypes.h>
#if 0
/* (c1*w1 + c2*w2) >> s */
inline __attribute__((always_inline)) uint32_t interp_2px(uint32_t c1, int w1, uint32_t c2, int w2, int s)
{
return (((((c1 & 0xff00ff00) >> 8) * w1 + ((c2 & 0xff00ff00) >> 8) * w2) << (8 - s)) & 0xff00ff00) |
(((((c1 & 0x00ff00ff) ) * w1 + ((c2 & 0x00ff00ff) ) * w2) >> s ) & 0x00ff00ff);
}
/* (c1*w1 + c2*w2 + c3*w3) >> s */
inline __attribute__((always_inline)) uint32_t interp_3px(uint32_t c1, int w1, uint32_t c2, int w2, uint32_t c3, int w3, int s)
{
return (((((c1 & 0xff00ff00) >> 8) * w1 + ((c2 & 0xff00ff00) >> 8) * w2 + ((c3 & 0xff00ff00) >> 8) * w3) << (8 - s)) & 0xff00ff00) |
(((((c1 & 0x00ff00ff) ) * w1 + ((c2 & 0x00ff00ff) ) * w2 + ((c3 & 0x00ff00ff) ) * w3) >> s ) & 0x00ff00ff);
}
#else
/* (c1*w1 + c2*w2) >> s */
inline __attribute__((always_inline)) uint32_t interp_2px(uint32_t c1, int w1, uint32_t c2, int w2, int s)
{
return w2 > w1 ? c2 : c1;
}
/* (c1*w1 + c2*w2 + c3*w3) >> s */
inline __attribute__((always_inline)) uint32_t interp_3px(uint32_t c1, int w1, uint32_t c2, int w2, uint32_t c3, int w3, int s)
{
if (w3 > w2 && w3 > w1)
return c3;
if (w2 > w1)
return c2;
return c1;
}
#endif
/* m is the mask of diff with the center pixel that matters in the pattern, and
* r is the expected result (bit set to 1 if there is difference with the
* center, 0 otherwise) */
#define P(m, r) ((k_shuffled & (m)) == (r))
/* adjust 012345678 to 01235678: the mask doesn't contain the (null) diff
* between the center/current pixel and itself */
#define DROP4(z) ((z) > 4 ? (z)-1 : (z))
/* shuffle the input mask: move bit n (4-adjusted) to position stored in p<n> */
#define SHF(x, rot, n) (((x) >> ((rot) ? 7-DROP4(n) : DROP4(n)) & 1) << DROP4(p##n))
/* used to check if there is YUV difference between 2 pixels */
#define WDIFF(c1, c2) (c1 != c2)
/* bootstrap template for every interpolation code. It defines the shuffled
* masks and surrounding pixels. The rot flag is used to indicate if it's a
* rotation; its basic effect is to shuffle k using p8..p0 instead of p0..p8 */
#define INTERP_BOOTSTRAP(rot) \
const int k_shuffled = SHF(k,rot,0) | SHF(k,rot,1) | SHF(k,rot,2) \
| SHF(k,rot,3) | 0 | SHF(k,rot,5) \
| SHF(k,rot,6) | SHF(k,rot,7) | SHF(k,rot,8); \
\
const uint32_t w0 = w[p0], w1 = w[p1], \
w3 = w[p3], w4 = w[p4], w5 = w[p5], \
w7 = w[p7]
/* Assuming p0..p8 is mapped to pixels 0..8, this function interpolates the
* top-left pixel in the total of the 2x2 pixels to interpolates. The function
* is also used for the 3 other pixels */
inline __attribute__((always_inline)) uint32_t hq2x_interp_1x1(int k,
const uint32_t *w,
int p0, int p1, int p2,
int p3, int p4, int p5,
int p6, int p7, int p8)
{
INTERP_BOOTSTRAP(0);
if ((P(0xbf,0x37) || P(0xdb,0x13)) && WDIFF(w1, w5))
return interp_2px(w4, 3, w3, 1, 2);
if ((P(0xdb,0x49) || P(0xef,0x6d)) && WDIFF(w7, w3))
return interp_2px(w4, 3, w1, 1, 2);
if ((P(0x0b,0x0b) || P(0xfe,0x4a) || P(0xfe,0x1a)) && WDIFF(w3, w1))
return w4;
if ((P(0x6f,0x2a) || P(0x5b,0x0a) || P(0xbf,0x3a) || P(0xdf,0x5a) ||
P(0x9f,0x8a) || P(0xcf,0x8a) || P(0xef,0x4e) || P(0x3f,0x0e) ||
P(0xfb,0x5a) || P(0xbb,0x8a) || P(0x7f,0x5a) || P(0xaf,0x8a) ||
P(0xeb,0x8a)) && WDIFF(w3, w1))
return interp_2px(w4, 3, w0, 1, 2);
if (P(0x0b,0x08))
return interp_3px(w4, 2, w0, 1, w1, 1, 2);
if (P(0x0b,0x02))
return interp_3px(w4, 2, w0, 1, w3, 1, 2);
if (P(0x2f,0x2f))
return interp_3px(w4, 14, w3, 1, w1, 1, 4);
if (P(0xbf,0x37) || P(0xdb,0x13))
return interp_3px(w4, 5, w1, 2, w3, 1, 3);
if (P(0xdb,0x49) || P(0xef,0x6d))
return interp_3px(w4, 5, w3, 2, w1, 1, 3);
if (P(0x1b,0x03) || P(0x4f,0x43) || P(0x8b,0x83) || P(0x6b,0x43))
return interp_2px(w4, 3, w3, 1, 2);
if (P(0x4b,0x09) || P(0x8b,0x89) || P(0x1f,0x19) || P(0x3b,0x19))
return interp_2px(w4, 3, w1, 1, 2);
if (P(0x7e,0x2a) || P(0xef,0xab) || P(0xbf,0x8f) || P(0x7e,0x0e))
return interp_3px(w4, 2, w3, 3, w1, 3, 3);
if (P(0xfb,0x6a) || P(0x6f,0x6e) || P(0x3f,0x3e) || P(0xfb,0xfa) ||
P(0xdf,0xde) || P(0xdf,0x1e))
return interp_2px(w4, 3, w0, 1, 2);
if (P(0x0a,0x00) || P(0x4f,0x4b) || P(0x9f,0x1b) || P(0x2f,0x0b) ||
P(0xbe,0x0a) || P(0xee,0x0a) || P(0x7e,0x0a) || P(0xeb,0x4b) ||
P(0x3b,0x1b))
return interp_3px(w4, 2, w3, 1, w1, 1, 2);
return interp_3px(w4, 6, w3, 1, w1, 1, 3);
}
/* Assuming p0..p8 is mapped to pixels 0..8, this function interpolates the
* top-left and top-center pixel in the total of the 3x3 pixels to
* interpolates. The function is also used for the 3 other couples of pixels
* defining the outline. The center pixel is not defined through this function,
* since it's just the same as the original value. */
inline __attribute__((always_inline)) void hq3x_interp_2x1(uint32_t *dst, int dst_linesize,
int k,
const uint32_t *w,
int pos00, int pos01,
int p0, int p1, int p2,
int p3, int p4, int p5,
int p6, int p7, int p8,
int rotate)
{
INTERP_BOOTSTRAP(rotate);
uint32_t *dst00 = &dst[dst_linesize*(pos00>>1) + (pos00&1)];
uint32_t *dst01 = &dst[dst_linesize*(pos01>>1) + (pos01&1)];
if ((P(0xdb,0x49) || P(0xef,0x6d)) && WDIFF(w7, w3))
*dst00 = interp_2px(w4, 3, w1, 1, 2);
else if ((P(0xbf,0x37) || P(0xdb,0x13)) && WDIFF(w1, w5))
*dst00 = interp_2px(w4, 3, w3, 1, 2);
else if ((P(0x0b,0x0b) || P(0xfe,0x4a) || P(0xfe,0x1a)) && WDIFF(w3, w1))
*dst00 = w4;
else if ((P(0x6f,0x2a) || P(0x5b,0x0a) || P(0xbf,0x3a) || P(0xdf,0x5a) ||
P(0x9f,0x8a) || P(0xcf,0x8a) || P(0xef,0x4e) || P(0x3f,0x0e) ||
P(0xfb,0x5a) || P(0xbb,0x8a) || P(0x7f,0x5a) || P(0xaf,0x8a) ||
P(0xeb,0x8a)) && WDIFF(w3, w1))
*dst00 = interp_2px(w4, 3, w0, 1, 2);
else if (P(0x4b,0x09) || P(0x8b,0x89) || P(0x1f,0x19) || P(0x3b,0x19))
*dst00 = interp_2px(w4, 3, w1, 1, 2);
else if (P(0x1b,0x03) || P(0x4f,0x43) || P(0x8b,0x83) || P(0x6b,0x43))
*dst00 = interp_2px(w4, 3, w3, 1, 2);
else if (P(0x7e,0x2a) || P(0xef,0xab) || P(0xbf,0x8f) || P(0x7e,0x0e))
*dst00 = interp_2px(w3, 1, w1, 1, 1);
else if (P(0x4f,0x4b) || P(0x9f,0x1b) || P(0x2f,0x0b) || P(0xbe,0x0a) ||
P(0xee,0x0a) || P(0x7e,0x0a) || P(0xeb,0x4b) || P(0x3b,0x1b))
*dst00 = interp_3px(w4, 2, w3, 7, w1, 7, 4);
else if (P(0x0b,0x08) || P(0xf9,0x68) || P(0xf3,0x62) || P(0x6d,0x6c) ||
P(0x67,0x66) || P(0x3d,0x3c) || P(0x37,0x36) || P(0xf9,0xf8) ||
P(0xdd,0xdc) || P(0xf3,0xf2) || P(0xd7,0xd6) || P(0xdd,0x1c) ||
P(0xd7,0x16) || P(0x0b,0x02))
*dst00 = interp_2px(w4, 3, w0, 1, 2);
else
*dst00 = interp_3px(w4, 2, w3, 1, w1, 1, 2);
if ((P(0xfe,0xde) || P(0x9e,0x16) || P(0xda,0x12) || P(0x17,0x16) ||
P(0x5b,0x12) || P(0xbb,0x12)) && WDIFF(w1, w5))
*dst01 = w4;
else if ((P(0x0f,0x0b) || P(0x5e,0x0a) || P(0xfb,0x7b) || P(0x3b,0x0b) ||
P(0xbe,0x0a) || P(0x7a,0x0a)) && WDIFF(w3, w1))
*dst01 = w4;
else if (P(0xbf,0x8f) || P(0x7e,0x0e) || P(0xbf,0x37) || P(0xdb,0x13))
*dst01 = interp_2px(w1, 3, w4, 1, 2);
else if (P(0x02,0x00) || P(0x7c,0x28) || P(0xed,0xa9) || P(0xf5,0xb4) ||
P(0xd9,0x90))
*dst01 = interp_2px(w4, 3, w1, 1, 2);
else if (P(0x4f,0x4b) || P(0xfb,0x7b) || P(0xfe,0x7e) || P(0x9f,0x1b) ||
P(0x2f,0x0b) || P(0xbe,0x0a) || P(0x7e,0x0a) || P(0xfb,0x4b) ||
P(0xfb,0xdb) || P(0xfe,0xde) || P(0xfe,0x56) || P(0x57,0x56) ||
P(0x97,0x16) || P(0x3f,0x1e) || P(0xdb,0x12) || P(0xbb,0x12))
*dst01 = interp_2px(w4, 7, w1, 1, 3);
else
*dst01 = w4;
}
/* Assuming p0..p8 is mapped to pixels 0..8, this function interpolates the
* top-left block of 2x2 pixels in the total of the 4x4 pixels (or 4 blocks) to
* interpolates. The function is also used for the 3 other blocks of 2x2
* pixels. */
inline __attribute__((always_inline)) void hq4x_interp_2x2(uint32_t *dst, int dst_linesize,
int k,
const uint32_t *w,
int pos00, int pos01,
int pos10, int pos11,
int p0, int p1, int p2,
int p3, int p4, int p5,
int p6, int p7, int p8)
{
INTERP_BOOTSTRAP(0);
uint32_t *dst00 = &dst[dst_linesize*(pos00>>1) + (pos00&1)];
uint32_t *dst01 = &dst[dst_linesize*(pos01>>1) + (pos01&1)];
uint32_t *dst10 = &dst[dst_linesize*(pos10>>1) + (pos10&1)];
uint32_t *dst11 = &dst[dst_linesize*(pos11>>1) + (pos11&1)];
const int cond00 = (P(0xbf,0x37) || P(0xdb,0x13)) && WDIFF(w1, w5);
const int cond01 = (P(0xdb,0x49) || P(0xef,0x6d)) && WDIFF(w7, w3);
const int cond02 = (P(0x6f,0x2a) || P(0x5b,0x0a) || P(0xbf,0x3a) ||
P(0xdf,0x5a) || P(0x9f,0x8a) || P(0xcf,0x8a) ||
P(0xef,0x4e) || P(0x3f,0x0e) || P(0xfb,0x5a) ||
P(0xbb,0x8a) || P(0x7f,0x5a) || P(0xaf,0x8a) ||
P(0xeb,0x8a)) && WDIFF(w3, w1);
const int cond03 = P(0xdb,0x49) || P(0xef,0x6d);
const int cond04 = P(0xbf,0x37) || P(0xdb,0x13);
const int cond05 = P(0x1b,0x03) || P(0x4f,0x43) || P(0x8b,0x83) ||
P(0x6b,0x43);
const int cond06 = P(0x4b,0x09) || P(0x8b,0x89) || P(0x1f,0x19) ||
P(0x3b,0x19);
const int cond07 = P(0x0b,0x08) || P(0xf9,0x68) || P(0xf3,0x62) ||
P(0x6d,0x6c) || P(0x67,0x66) || P(0x3d,0x3c) ||
P(0x37,0x36) || P(0xf9,0xf8) || P(0xdd,0xdc) ||
P(0xf3,0xf2) || P(0xd7,0xd6) || P(0xdd,0x1c) ||
P(0xd7,0x16) || P(0x0b,0x02);
const int cond08 = (P(0x0f,0x0b) || P(0x2b,0x0b) || P(0xfe,0x4a) ||
P(0xfe,0x1a)) && WDIFF(w3, w1);
const int cond09 = P(0x2f,0x2f);
const int cond10 = P(0x0a,0x00);
const int cond11 = P(0x0b,0x09);
const int cond12 = P(0x7e,0x2a) || P(0xef,0xab);
const int cond13 = P(0xbf,0x8f) || P(0x7e,0x0e);
const int cond14 = P(0x4f,0x4b) || P(0x9f,0x1b) || P(0x2f,0x0b) ||
P(0xbe,0x0a) || P(0xee,0x0a) || P(0x7e,0x0a) ||
P(0xeb,0x4b) || P(0x3b,0x1b);
const int cond15 = P(0x0b,0x03);
if (cond00)
*dst00 = interp_2px(w4, 5, w3, 3, 3);
else if (cond01)
*dst00 = interp_2px(w4, 5, w1, 3, 3);
else if ((P(0x0b,0x0b) || P(0xfe,0x4a) || P(0xfe,0x1a)) && WDIFF(w3, w1))
*dst00 = w4;
else if (cond02)
*dst00 = interp_2px(w4, 5, w0, 3, 3);
else if (cond03)
*dst00 = interp_2px(w4, 3, w3, 1, 2);
else if (cond04)
*dst00 = interp_2px(w4, 3, w1, 1, 2);
else if (cond05)
*dst00 = interp_2px(w4, 5, w3, 3, 3);
else if (cond06)
*dst00 = interp_2px(w4, 5, w1, 3, 3);
else if (P(0x0f,0x0b) || P(0x5e,0x0a) || P(0x2b,0x0b) || P(0xbe,0x0a) ||
P(0x7a,0x0a) || P(0xee,0x0a))
*dst00 = interp_2px(w1, 1, w3, 1, 1);
else if (cond07)
*dst00 = interp_2px(w4, 5, w0, 3, 3);
else
*dst00 = interp_3px(w4, 2, w1, 1, w3, 1, 2);
if (cond00)
*dst01 = interp_2px(w4, 7, w3, 1, 3);
else if (cond08)
*dst01 = w4;
else if (cond02)
*dst01 = interp_2px(w4, 3, w0, 1, 2);
else if (cond09)
*dst01 = w4;
else if (cond10)
*dst01 = interp_3px(w4, 5, w1, 2, w3, 1, 3);
else if (P(0x0b,0x08))
*dst01 = interp_3px(w4, 5, w1, 2, w0, 1, 3);
else if (cond11)
*dst01 = interp_2px(w4, 5, w1, 3, 3);
else if (cond04)
*dst01 = interp_2px(w1, 3, w4, 1, 2);
else if (cond12)
*dst01 = interp_3px(w1, 2, w4, 1, w3, 1, 2);
else if (cond13)
*dst01 = interp_2px(w1, 5, w3, 3, 3);
else if (cond05)
*dst01 = interp_2px(w4, 7, w3, 1, 3);
else if (P(0xf3,0x62) || P(0x67,0x66) || P(0x37,0x36) || P(0xf3,0xf2) ||
P(0xd7,0xd6) || P(0xd7,0x16) || P(0x0b,0x02))
*dst01 = interp_2px(w4, 3, w0, 1, 2);
else if (cond14)
*dst01 = interp_2px(w1, 1, w4, 1, 1);
else
*dst01 = interp_2px(w4, 3, w1, 1, 2);
if (cond01)
*dst10 = interp_2px(w4, 7, w1, 1, 3);
else if (cond08)
*dst10 = w4;
else if (cond02)
*dst10 = interp_2px(w4, 3, w0, 1, 2);
else if (cond09)
*dst10 = w4;
else if (cond10)
*dst10 = interp_3px(w4, 5, w3, 2, w1, 1, 3);
else if (P(0x0b,0x02))
*dst10 = interp_3px(w4, 5, w3, 2, w0, 1, 3);
else if (cond15)
*dst10 = interp_2px(w4, 5, w3, 3, 3);
else if (cond03)
*dst10 = interp_2px(w3, 3, w4, 1, 2);
else if (cond13)
*dst10 = interp_3px(w3, 2, w4, 1, w1, 1, 2);
else if (cond12)
*dst10 = interp_2px(w3, 5, w1, 3, 3);
else if (cond06)
*dst10 = interp_2px(w4, 7, w1, 1, 3);
else if (P(0x0b,0x08) || P(0xf9,0x68) || P(0x6d,0x6c) || P(0x3d,0x3c) ||
P(0xf9,0xf8) || P(0xdd,0xdc) || P(0xdd,0x1c))
*dst10 = interp_2px(w4, 3, w0, 1, 2);
else if (cond14)
*dst10 = interp_2px(w3, 1, w4, 1, 1);
else
*dst10 = interp_2px(w4, 3, w3, 1, 2);
if ((P(0x7f,0x2b) || P(0xef,0xab) || P(0xbf,0x8f) || P(0x7f,0x0f)) &&
WDIFF(w3, w1))
*dst11 = w4;
else if (cond02)
*dst11 = interp_2px(w4, 7, w0, 1, 3);
else if (cond15)
*dst11 = interp_2px(w4, 7, w3, 1, 3);
else if (cond11)
*dst11 = interp_2px(w4, 7, w1, 1, 3);
else if (P(0x0a,0x00) || P(0x7e,0x2a) || P(0xef,0xab) || P(0xbf,0x8f) ||
P(0x7e,0x0e))
*dst11 = interp_3px(w4, 6, w3, 1, w1, 1, 3);
else if (cond07)
*dst11 = interp_2px(w4, 7, w0, 1, 3);
else
*dst11 = w4;
}
void hqx_filter(uint32_t* src, uint32_t* dst, int width, int height, int n)
{
int x, y;
const int dst32_linesize = width * n;
const int src32_linesize = width;
uint32_t *src32, *dst32;
src32 = src;
dst32 = dst;
for (y = 0; y < height; y++) {
const int prevline = y > 0 ? -src32_linesize : 0;
const int nextline = y < height - 1 ? src32_linesize : 0;
for (x = 0; x < width; x++) {
const int prevcol = x > 0 ? -1 : 0;
const int nextcol = x < width -1 ? 1 : 0;
const uint32_t w[3*3] = {
src32[prevcol + prevline], src32[prevline], src32[prevline + nextcol],
src32[prevcol ], src32[ 0], src32[ nextcol],
src32[prevcol + nextline], src32[nextline], src32[nextline + nextcol]
};
const uint32_t yuv1 = w[4];
const int pattern = (w[4] != w[0] ? 1 : 0)
| (w[4] != w[1] ? 1 : 0) << 1
| (w[4] != w[2] ? 1 : 0) << 2
| (w[4] != w[3] ? 1 : 0) << 3
| (w[4] != w[5] ? 1 : 0) << 4
| (w[4] != w[6] ? 1 : 0) << 5
| (w[4] != w[7] ? 1 : 0) << 6
| (w[4] != w[8] ? 1 : 0) << 7;
if (n == 2) {
dst32[dst32_linesize*0 + 0] = hq2x_interp_1x1(pattern, w, 0,1,2,3,4,5,6,7,8); // 00
dst32[dst32_linesize*0 + 1] = hq2x_interp_1x1(pattern, w, 2,1,0,5,4,3,8,7,6); // 01 (vert mirrored)
dst32[dst32_linesize*1 + 0] = hq2x_interp_1x1(pattern, w, 6,7,8,3,4,5,0,1,2); // 10 (horiz mirrored)
dst32[dst32_linesize*1 + 1] = hq2x_interp_1x1(pattern, w, 8,7,6,5,4,3,2,1,0); // 11 (center mirrored)
} else if (n == 3) {
hq3x_interp_2x1(dst32, dst32_linesize, pattern, w, 0,1, 0,1,2,3,4,5,6,7,8, 0); // 00 01
hq3x_interp_2x1(dst32 + 1, dst32_linesize, pattern, w, 1,3, 2,5,8,1,4,7,0,3,6, 1); // 02 12 (rotated to the right)
hq3x_interp_2x1(dst32 + 1*dst32_linesize, dst32_linesize, pattern, w, 2,0, 6,3,0,7,4,1,8,5,2, 1); // 20 10 (rotated to the left)
hq3x_interp_2x1(dst32 + 1*dst32_linesize + 1, dst32_linesize, pattern, w, 3,2, 8,7,6,5,4,3,2,1,0, 0); // 22 21 (center mirrored)
dst32[dst32_linesize + 1] = w[4]; // 11
} else if (n == 4) {
hq4x_interp_2x2(dst32, dst32_linesize, pattern, w, 0,1,2,3, 0,1,2,3,4,5,6,7,8); // 00 01 10 11
hq4x_interp_2x2(dst32 + 2, dst32_linesize, pattern, w, 1,0,3,2, 2,1,0,5,4,3,8,7,6); // 02 03 12 13 (vert mirrored)
hq4x_interp_2x2(dst32 + 2*dst32_linesize, dst32_linesize, pattern, w, 2,3,0,1, 6,7,8,3,4,5,0,1,2); // 20 21 30 31 (horiz mirrored)
hq4x_interp_2x2(dst32 + 2*dst32_linesize + 2, dst32_linesize, pattern, w, 3,2,1,0, 8,7,6,5,4,3,2,1,0); // 22 23 32 33 (center mirrored)
} else {
return;
}
src32 += 1;
dst32 += n;
}
dst32 += dst32_linesize * (n - 1);
}
}
#include <inttypes.h>
#ifdef __cplusplus
extern "C" {
#endif
void hqx_filter(uint32_t* src, uint32_t* dst, int height, int width, int n);
#ifdef __cplusplus
}
#endif
......@@ -4,5 +4,6 @@ OBJS = $(MTOBJODIR)$(DIRSEP)ansi_cio$(OFILE)\
$(MTOBJODIR)$(DIRSEP)vidmodes$(OFILE)\
$(MTOBJODIR)$(DIRSEP)allfonts$(OFILE)\
$(MTOBJODIR)$(DIRSEP)mouse$(OFILE)\
$(MTOBJODIR)$(DIRSEP)scale$(OFILE)\
$(MTOBJODIR)$(DIRSEP)syncicon64$(OFILE)\
$(MTOBJODIR)$(DIRSEP)utf8_codepages$(OFILE)
#include "scale.h"
#include "xbr.h"
static void pointy_scale3(uint32_t* src, uint32_t* dest, int width, int height);
static void pointy_scale5(uint32_t* src, uint32_t* dest, int width, int height);
uint32_t*
do_scale(struct rectlist* rect, uint32_t **target, size_t *targetsz, int* xscale, int* yscale, int* w, int* h)
{
int pointy5 = 0;
int pointy3 = 0;
int xbr2 = 0;
int xbr4 = 0;
int newscale = 1;
int total_scaling = 1;
static uint32_t *tmptarget = NULL;
static size_t tmptargetsz = 0;
uint32_t *ctarget;
uint32_t *csrc = rect->data;
uint32_t* nt;
int swidth = rect->rect.width;
int sheight = rect->rect.height;
*w = rect->rect.width;
*h = rect->rect.height;
// As a first pass, only do equal integer scaling
if (*xscale != *yscale)
return rect->data;
switch (*xscale) {
case 1:
return rect->data;
case 2:
xbr2 = 1;
break;
case 3:
pointy3 = 1;
break;
case 4:
xbr4 = 1;
break;
case 5:
pointy5 = 1;
break;
case 6:
pointy3 = 1;
xbr2 = 1;
break;
default:
total_scaling = *xscale;
*xscale = 1;
while (total_scaling > 1 && ((total_scaling % 5) == 0)) {
pointy5++;
total_scaling /= 5;
*xscale *= 5;
}
while (total_scaling > 1 && ((total_scaling % 3) == 0)) {
pointy3++;
total_scaling /= 3;
*xscale *= 3;
}
while (total_scaling > 1 && ((total_scaling % 4) == 0)) {
xbr4++;
total_scaling /= 4;
*xscale *= 4;
}
while (total_scaling > 1 && ((total_scaling % 2) == 0)) {
xbr2++;
total_scaling /= 2;
*xscale *= 2;
}
if (*xscale == 1)
return rect->data;
break;
}
// Now make sure target is big enough...
size_t needsz = rect->rect.width * rect->rect.height * (*xscale) * (*xscale) * sizeof(uint32_t);
if (needsz > *targetsz) {
nt = realloc(*target, needsz);
if (nt == NULL)
return rect->data;
*target = nt;
*targetsz = needsz;
}
ctarget = *target;
// And if we need an extra target, do the same there...
if (pointy3 + pointy5 + xbr4 + xbr2 > 1) {
if (needsz > tmptargetsz) {
nt = realloc(tmptarget, needsz);
if (nt == NULL)
return rect->data;
tmptarget = nt;
tmptargetsz = needsz;
}
}
// And finally, scale...
while (xbr4 > 0) {
xbr_filter(csrc, ctarget, swidth, sheight, 4);
xbr4--;
swidth *= 4;
sheight *= 4;
csrc = ctarget;
if (ctarget == tmptarget)
ctarget = *target;
else
ctarget = tmptarget;
}
while (xbr2 > 0) {
xbr_filter(csrc, ctarget, swidth, sheight, 2);
xbr2--;
swidth *= 2;
sheight *= 2;
csrc = ctarget;
if (ctarget == tmptarget)
ctarget = *target;
else
ctarget = tmptarget;
}
while (pointy5 > 0) {
pointy_scale5(csrc, ctarget, swidth, sheight);
pointy5--;
swidth *= 5;
sheight *= 5;
csrc = ctarget;
if (ctarget == tmptarget)
ctarget = *target;
else
ctarget = tmptarget;
}
while (pointy3 > 0) {
pointy_scale3(csrc, ctarget, swidth, sheight);
pointy3--;
swidth *= 3;
sheight *= 3;
csrc = ctarget;
if (ctarget == tmptarget)
ctarget = *target;
else
ctarget = tmptarget;
}
*w *= *xscale;
*h *= *xscale;
*xscale = newscale;
*yscale = newscale;
return csrc;
}
static void
pointy_scale5(uint32_t* src, uint32_t* dest, int width, int height)
{
int x, y;
uint32_t* s;
uint32_t* d;
int w5 = width * 5;
int w10 = width * 10;
int w15 = width * 15;
int w20 = width * 20;
int prevline, prevcol, nextline, nextcol;
s = src;
d = dest;
prevline = 0;
nextline = width;
for (y = 0; y < height; y++) {
if (y == height - 1)
nextline = 0;
prevcol = 0;
nextcol = 1;
for (x = 0; x < width; x++) {
if (x == width - 1)
nextcol = 0;
if (s[prevline + prevcol] == s[0]) {
d[0] = s[0];
d[1] = s[0];
d[w5] = s[0];
}
else if (s[prevcol] == s[prevline]) {
d[0] = s[prevcol];
d[1] = s[prevcol];
d[w5] = s[prevcol];
}
else {
d[0] = *s;
d[1] = *s;
d[w5] = *s;
}
// Top-middle stays OG.
d[2] = *s;
d[w5+1] = *s;
d[w5+2] = *s;
d[w5+3] = *s;
// And so on around the outside (round the outside)
if (s[prevline + nextcol] == s[0]) {
d[3] = s[0];
d[4] = s[0];
d[w5 + 4] = s[0];
}
else if (s[nextcol] == s[prevline]) {
d[3] = s[nextcol];
d[4] = s[nextcol];
d[w5 + 4] = s[nextcol];
}
else {
d[3] = s[0];
d[4] = s[0];
d[w5 + 4] = s[0];
}
d[w10] = *s;
d[w10+1] = *s;
d[w10+2] = *s;
d[w10+3] = *s;
d[w10+4] = *s;
if (s[prevcol + nextline] == s[0]) {
d[w15] = s[0];
d[w20] = s[0];
d[w20 + 1] = s[0];
}
else if(s[prevcol] == s[nextline]) {
d[w15] = s[prevcol];
d[w20] = s[prevcol];
d[w20 + 1] = s[prevcol];
}
else {
d[w15] = s[0];
d[w20] = s[0];
d[w20 + 1] = s[0];
}
d[w15 + 1] = *s;
d[w15 + 2] = *s;
d[w15 + 3] = *s;
d[w20 + 2] = *s;
if (s[nextcol + nextline] == s[0]) {
d[w15 + 4] = s[0];
d[w20 + 3] = s[0];
d[w20 + 4] = s[0];
}
else if (s[nextcol] == s[nextline]) {
d[w15 + 4] = s[nextcol];
d[w20 + 3] = s[nextcol];
d[w20 + 4] = s[nextcol];
}
else {
d[w15 + 4] = s[0];
d[w20 + 3] = s[0];
d[w20 + 4] = s[0];
}
d += 5;
s++;
if (x == 0)
prevcol = -1;
}
d += w20;
if (y == 0)
prevline = -width;
}
}
static void
pointy_scale3(uint32_t* src, uint32_t* dest, int width, int height)
{
int x, y;
uint32_t* s;
uint32_t* d;
int w3 = width * 3;
int w6 = width * 6;
int prevline, prevcol, nextline, nextcol;
s = src;
d = dest;
prevline = 0;
nextline = width;
for (y = 0; y < height; y++) {
if (y == height - 1)
nextline = 0;
prevcol = 0;
nextcol = 1;
for (x = 0; x < width; x++) {
if (x == width - 1)
nextcol = 0;
// Top-left is filled if both left and top are the same.
if (s[prevline + prevcol] == s[0])
d[0] = s[0];
else if (s[prevcol] == s[prevline])
d[0] = s[prevcol];
else
d[0] = *s;
// Top-middle stays OG.
d[1] = *s;
// And so on around the outside (round the outside)
if (s[prevline + nextcol] == s[0])
d[2] = s[0];
else if (s[nextcol] == s[prevline])
d[2] = s[nextcol];
else
d[2] = *s;
d[w3] = *s;
d[w3 + 1] = *s;
d[w3 + 2] = *s;
if (s[prevcol + nextline] == s[0])
d[w6] = s[0];
else if(s[prevcol] == s[nextline])
d[w6] = s[prevcol];
else
d[w6] = *s;
d[w6 + 1] = *s;
if (s[nextcol + nextline] == s[0])
d[w6 + 2] = s[0];
else if (s[nextcol] == s[nextline])
d[w6 + 2] = s[nextcol];
else
d[w6 + 2] = s[0];
d += 3;
s++;
if (x == 0)
prevcol = -1;
}
d += w6;
if (y == 0)
prevline = -width;
}
}
#include "bitmap_con.h"
uint32_t* do_scale(struct rectlist* rect, uint32_t **target, size_t *targetsz, int* xscale, int* yscale, int* w, int* h);
......@@ -27,6 +27,7 @@
#define BITMAP_CIOLIB_DRIVER
#include "bitmap_con.h"
#include "link_list.h"
#include "scale.h"
#include "x_events.h"
#include "x_cio.h"
#include "utf8_codepages.h"
......@@ -449,16 +450,25 @@ static int video_init()
static void local_draw_rect(struct rectlist *rect)
{
int x,y,xscale,yscale,xoff=0,yoff=0;
int xscaling, yscaling;
int rxscaling, ryscaling;
unsigned int r, g, b;
unsigned long pixel;
int cleft = rect->rect.width;
int cright = -1;
int ctop = rect->rect.height;
int cleft;
int cright = -100;
int ctop;
int cbottom = -1;
int idx;
int ridx;
int ridx_part;
uint32_t last_pixel = 0x55555555;
static uint32_t* target = NULL;
size_t targetsz = 0;
uint32_t* source;
int width, height;
int lheight, lines;
if (bitmap_width != cleft || bitmap_height != ctop)
if (bitmap_width != rect->rect.width || bitmap_height != rect->rect.height)
return;
xoff = (x11_window_width - xim->width) / 2;
......@@ -468,12 +478,26 @@ static void local_draw_rect(struct rectlist *rect)
if (yoff < 0)
yoff = 0;
// Scale...
xscaling = x_cvstat.scaling;
yscaling = x_cvstat.scaling*x_cvstat.vmultiplier;
source = do_scale(rect, &target, &targetsz, &xscaling, &yscaling, &width, &height);
rxscaling = width / rect->rect.width;
ryscaling = height / rect->rect.height;
cleft = width;
ctop = height;
lines = 0;
lheight = x_cvstat.charheight * ryscaling;
/* TODO: Translate into local colour depth */
for(y=0;y<rect->rect.height;y++) {
idx = y*rect->rect.width;
for(x=0; x<rect->rect.width; x++) {
for(y=0;y<height;y++) {
idx = y * width;
ridx = y / ryscaling * rect->rect.width;
ridx_part = 0;
for(x=0; x<width; x++) {
if (last) {
if (last->data[idx] != rect->data[idx]) {
// TODO: Based on source pixel, not target pixel. :(
if (last->data[ridx] != rect->data[ridx]) {
if (x < cleft)
cleft = x;
if (x > cright)
......@@ -484,15 +508,29 @@ static void local_draw_rect(struct rectlist *rect)
cbottom = y;
}
else {
idx++;
continue;
if (source == rect->data) {
idx += (rxscaling);
x += (rxscaling - 1);
ridx++;
continue;
}
else {
if (cright < x - rxscaling * 2
&& (x >= width - (rxscaling * 3)
|| last->data[ridx + 1] == rect->data[ridx + 1])) {
idx += (rxscaling);
x += (rxscaling - 1);
ridx++;
continue;
}
}
}
}
if (last_pixel != rect->data[idx]) {
last_pixel = rect->data[idx];
r = rect->data[idx] >> 16 & 0xff;
g = rect->data[idx] >> 8 & 0xff;
b = rect->data[idx] & 0xff;
if (last_pixel != source[idx]) {
last_pixel = source[idx];
r = source[idx] >> 16 & 0xff;
g = source[idx] >> 8 & 0xff;
b = source[idx] & 0xff;
r = (r<<8)|r;
g = (g<<8)|g;
b = (b<<8)|b;
......@@ -510,28 +548,49 @@ static void local_draw_rect(struct rectlist *rect)
else
pixel |= (b >> (0-b_shift)) & visual.blue_mask;
}
for(yscale=0; yscale<x_cvstat.scaling*x_cvstat.vmultiplier; yscale++) {
for(xscale=0; xscale<x_cvstat.scaling; xscale++) {
for(yscale=0; yscale<yscaling; yscale++) {
for(xscale=0; xscale<xscaling; xscale++) {
#ifdef XPutPixel
XPutPixel(xim,(x+rect->rect.x)*x_cvstat.scaling+xscale,(y+rect->rect.y)*x_cvstat.scaling*x_cvstat.vmultiplier+yscale,pixel);
XPutPixel(xim, (x + rect->rect.x) * xscaling + xscale
, (y + rect->rect.y) * yscaling + yscale, pixel);
#else
x11.XPutPixel(xim,(x+rect->rect.x)*x_cvstat.scaling+xscale,(y+rect->rect.y)*x_cvstat.scaling*x_cvstat.vmultiplier+yscale,pixel);
x11.XPutPixel(xim, (x + rect->rect.x) * xscaling + xscale
, (y + rect->rect.y) * yscaling + yscale, pixel);
#endif
}
}
idx++;
ridx_part++;
if (ridx_part >= rxscaling) {
ridx_part = 0;
ridx++;
}
}
lines++;
/* This line was changed */
if (last && (((y & 0x1f) == 0x1f) || (y == rect->rect.height-1)) && cright >= 0) {
x11.XPutImage(dpy, win, gc, xim, cleft*x_cvstat.scaling, ctop*x_cvstat.scaling*x_cvstat.vmultiplier, cleft*x_cvstat.scaling + xoff, ctop*x_cvstat.scaling*x_cvstat.vmultiplier + yoff, (cright-cleft+1)*x_cvstat.scaling, (cbottom-ctop+1)*x_cvstat.scaling*x_cvstat.vmultiplier);
cleft = rect->rect.width;
cright = cbottom = -1;
ctop = rect->rect.height;
if (last && ((lines == lheight) || (y == height - 1)) && cright >= 0) {
#ifdef USE_XBRZ
if (source != rect->data) {
cleft -= 2 * rxscaling;
if (cleft < 0)
cleft = 0;
cright += 2 * rxscaling;
if (cright >= width)
cright = width - 1;
}
#endif
lines = 0;
x11.XPutImage(dpy, win, gc, xim, cleft * xscaling, ctop * yscaling
, cleft * xscaling + xoff, ctop * yscaling + yoff
, (cright - cleft + 1) * xscaling, (cbottom - ctop + 1) * yscaling);
cleft = width;
cright = cbottom = -100;
ctop = height;
}
}
if (last == NULL)
x11.XPutImage(dpy, win, gc, xim, rect->rect.x*x_cvstat.scaling, rect->rect.y*x_cvstat.scaling*x_cvstat.vmultiplier, rect->rect.x*x_cvstat.scaling + xoff, rect->rect.y*x_cvstat.scaling*x_cvstat.vmultiplier + yoff, rect->rect.width*x_cvstat.scaling, rect->rect.height*x_cvstat.scaling*x_cvstat.vmultiplier);
x11.XPutImage(dpy, win, gc, xim, rect->rect.x*xscaling, rect->rect.y*yscaling, rect->rect.x*xscaling + xoff, rect->rect.y*yscaling + yoff, width * xscaling, height * yscaling);
else
bitmap_drv_free_rect(last);
last = rect;
......@@ -644,11 +703,14 @@ static int x11_event(XEvent *ev)
break;
/* Graphics related events */
case ConfigureNotify:
x11_window_xpos=ev->xconfigure.x;
x11_window_ypos=ev->xconfigure.y;
x11_window_width=ev->xconfigure.width;
x11_window_height=ev->xconfigure.height;
handle_resize_event(ev->xconfigure.width, ev->xconfigure.height);
if (x11_window_xpos != ev->xconfigure.x || x11_window_ypos != ev->xconfigure.y
|| x11_window_width != ev->xconfigure.width || x11_window_height != ev->xconfigure.height) {
x11_window_xpos=ev->xconfigure.x;
x11_window_ypos=ev->xconfigure.y;
x11_window_width=ev->xconfigure.width;
x11_window_height=ev->xconfigure.height;
handle_resize_event(ev->xconfigure.width, ev->xconfigure.height);
}
break;
case NoExpose:
break;
......
/*
* This file is part of FFmpeg.
*
* Copyright (c) 2011, 2012 Hyllian/Jararaca <sergiogdb@gmail.com>
* Copyright (c) 2014 Arwa Arif <arwaarif1994@gmail.com>
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* XBR Filter is used for depixelization of image.
* This is based on Hyllian's xBR shader.
*
* @see https://forums.libretro.com/t/xbr-algorithm-tutorial/123
* @see https://github.com/yoyofr/iFBA/blob/master/fba_src/src/intf/video/scalers/xbr.cpp
*/
#include <inttypes.h>
#include <stdlib.h>
#define LB_MASK 0x00FEFEFE
#define RED_BLUE_MASK 0x00FF00FF
#define GREEN_MASK 0x0000FF00
#ifdef PI
#undef PI
#endif
uint32_t r2y[1<<24];
static uint32_t pixel_diff(uint32_t x, uint32_t y)
{
#define YMASK 0xff0000
#define UMASK 0x00ff00
#define VMASK 0x0000ff
#define ABSDIFF(a,b) (abs((int)(a)-(int)(b)))
uint32_t yuv1 = r2y[x & 0xffffff];
uint32_t yuv2 = r2y[y & 0xffffff];
return (ABSDIFF(yuv1 & YMASK, yuv2 & YMASK) >> 16) +
(ABSDIFF(yuv1 & UMASK, yuv2 & UMASK) >> 8) +
ABSDIFF(yuv1 & VMASK, yuv2 & VMASK);
}
#define ALPHA_BLEND_128_W(a, b) ((((a) & LB_MASK) >> 1) + (((b) & LB_MASK) >> 1))
#define ALPHA_BLEND_BASE(a, b, m, s) ( (RED_BLUE_MASK & (((a) & RED_BLUE_MASK) + (((((b) & RED_BLUE_MASK) - ((a) & RED_BLUE_MASK)) * (m)) >> (s)))) \
| (GREEN_MASK & (((a) & GREEN_MASK) + (((((b) & GREEN_MASK) - ((a) & GREEN_MASK)) * (m)) >> (s)))))
#define ALPHA_BLEND_32_W(a, b) ALPHA_BLEND_BASE(a, b, 1, 3)
#define ALPHA_BLEND_64_W(a, b) ALPHA_BLEND_BASE(a, b, 1, 2)
#define ALPHA_BLEND_192_W(a, b) ALPHA_BLEND_BASE(a, b, 3, 2)
#define ALPHA_BLEND_224_W(a, b) ALPHA_BLEND_BASE(a, b, 7, 3)
#define df(A, B) pixel_diff(A, B)
#define eq(A, B) (df(A, B) < 155)
#define FILT2(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, \
N0, N1, N2, N3) do { \
if (PE != PH && PE != PF) { \
const unsigned e = df(PE,PC) + df(PE,PG) + df(PI,H5) + df(PI,F4) + (df(PH,PF)<<2); \
const unsigned i = df(PH,PD) + df(PH,I5) + df(PF,I4) + df(PF,PB) + (df(PE,PI)<<2); \
if (e <= i) { \
const unsigned px = df(PE,PF) <= df(PE,PH) ? PF : PH; \
if (e < i && (!eq(PF,PB) && !eq(PH,PD) || eq(PE,PI) \
&& (!eq(PF,I4) && !eq(PH,I5)) \
|| eq(PE,PG) || eq(PE,PC))) { \
const unsigned ke = df(PF,PG); \
const unsigned ki = df(PH,PC); \
const int left = ke<<1 <= ki && PE != PG && PD != PG; \
const int up = ke >= ki<<1 && PE != PC && PB != PC; \
if (left && up) { \
E[N3] = ALPHA_BLEND_224_W(E[N3], px); \
E[N2] = ALPHA_BLEND_64_W( E[N2], px); \
E[N1] = E[N2]; \
} else if (left) { \
E[N3] = ALPHA_BLEND_192_W(E[N3], px); \
E[N2] = ALPHA_BLEND_64_W( E[N2], px); \
} else if (up) { \
E[N3] = ALPHA_BLEND_192_W(E[N3], px); \
E[N1] = ALPHA_BLEND_64_W( E[N1], px); \
} else { /* diagonal */ \
E[N3] = ALPHA_BLEND_128_W(E[N3], px); \
} \
} else { \
E[N3] = ALPHA_BLEND_128_W(E[N3], px); \
} \
} \
} \
} while (0)
#define FILT3(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, \
N0, N1, N2, N3, N4, N5, N6, N7, N8) do { \
if (PE != PH && PE != PF) { \
const unsigned e = df(PE,PC) + df(PE,PG) + df(PI,H5) + df(PI,F4) + (df(PH,PF)<<2); \
const unsigned i = df(PH,PD) + df(PH,I5) + df(PF,I4) + df(PF,PB) + (df(PE,PI)<<2); \
if (e <= i) { \
const unsigned px = df(PE,PF) <= df(PE,PH) ? PF : PH; \
if (e < i && (!eq(PF,PB) && !eq(PF,PC) || !eq(PH,PD) && !eq(PH,PG) || eq(PE,PI) \
&& (!eq(PF,F4) && !eq(PF,I4) || !eq(PH,H5) && !eq(PH,I5)) \
|| eq(PE,PG) || eq(PE,PC))) { \
const unsigned ke = df(PF,PG); \
const unsigned ki = df(PH,PC); \
const int left = ke<<1 <= ki && PE != PG && PD != PG; \
const int up = ke >= ki<<1 && PE != PC && PB != PC; \
if (left && up) { \
E[N7] = ALPHA_BLEND_192_W(E[N7], px); \
E[N6] = ALPHA_BLEND_64_W( E[N6], px); \
E[N5] = E[N7]; \
E[N2] = E[N6]; \
E[N8] = px; \
} else if (left) { \
E[N7] = ALPHA_BLEND_192_W(E[N7], px); \
E[N5] = ALPHA_BLEND_64_W( E[N5], px); \
E[N6] = ALPHA_BLEND_64_W( E[N6], px); \
E[N8] = px; \
} else if (up) { \
E[N5] = ALPHA_BLEND_192_W(E[N5], px); \
E[N7] = ALPHA_BLEND_64_W( E[N7], px); \
E[N2] = ALPHA_BLEND_64_W( E[N2], px); \
E[N8] = px; \
} else { /* diagonal */ \
E[N8] = ALPHA_BLEND_224_W(E[N8], px); \
E[N5] = ALPHA_BLEND_32_W( E[N5], px); \
E[N7] = ALPHA_BLEND_32_W( E[N7], px); \
} \
} else { \
E[N8] = ALPHA_BLEND_128_W(E[N8], px); \
} \
} \
} \
} while (0)
#define FILT4(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, \
N15, N14, N11, N3, N7, N10, N13, N12, N9, N6, N2, N1, N5, N8, N4, N0) do { \
if (PE != PH && PE != PF) { \
const unsigned e = df(PE,PC) + df(PE,PG) + df(PI,H5) + df(PI,F4) + (df(PH,PF)<<2); \
const unsigned i = df(PH,PD) + df(PH,I5) + df(PF,I4) + df(PF,PB) + (df(PE,PI)<<2); \
if (e <= i) { \
const unsigned px = df(PE,PF) <= df(PE,PH) ? PF : PH; \
if (e < i && (!eq(PF,PB) && !eq(PH,PD) || eq(PE,PI) \
&& (!eq(PF,I4) && !eq(PH,I5)) \
|| eq(PE,PG) || eq(PE,PC))) { \
const unsigned ke = df(PF,PG); \
const unsigned ki = df(PH,PC); \
const int left = ke<<1 <= ki && PE != PG && PD != PG; \
const int up = ke >= ki<<1 && PE != PC && PB != PC; \
if (left && up) { \
E[N13] = ALPHA_BLEND_192_W(E[N13], px); \
E[N12] = ALPHA_BLEND_64_W( E[N12], px); \
E[N15] = E[N14] = E[N11] = px; \
E[N10] = E[N3] = E[N12]; \
E[N7] = E[N13]; \
} else if (left) { \
E[N11] = ALPHA_BLEND_192_W(E[N11], px); \
E[N13] = ALPHA_BLEND_192_W(E[N13], px); \
E[N10] = ALPHA_BLEND_64_W( E[N10], px); \
E[N12] = ALPHA_BLEND_64_W( E[N12], px); \
E[N14] = px; \
E[N15] = px; \
} else if (up) { \
E[N14] = ALPHA_BLEND_192_W(E[N14], px); \
E[N7 ] = ALPHA_BLEND_192_W(E[N7 ], px); \
E[N10] = ALPHA_BLEND_64_W( E[N10], px); \
E[N3 ] = ALPHA_BLEND_64_W( E[N3 ], px); \
E[N11] = px; \
E[N15] = px; \
} else { /* diagonal */ \
E[N11] = ALPHA_BLEND_128_W(E[N11], px); \
E[N14] = ALPHA_BLEND_128_W(E[N14], px); \
E[N15] = px; \
} \
} else { \
E[N15] = ALPHA_BLEND_128_W(E[N15], px); \
} \
} \
} \
} while (0)
void
xbr_filter(uint32_t *data, uint32_t *out, int width, int height, int n)
{
int x, y;
const int nl = width * n;
const int nl1 = nl + nl;
const int nl2 = nl1 + nl;
for (y = 0; y < height; y++) {
uint32_t *E = out + y * width * n * n;
const uint32_t *sa2 = (data + y * width - 2); /* center */
const uint32_t *sa1 = sa2 - (width); /* up x1 */
const uint32_t *sa0 = sa1 - (width); /* up x2 */
const uint32_t *sa3 = sa2 + (width); /* down x1 */
const uint32_t *sa4 = sa3 + (width); /* down x2 */
if (y <= 1) {
sa0 = sa1;
if (y == 0) {
sa0 = sa1 = sa2;
}
}
if (y >= height - 2) {
sa4 = sa3;
if (y == height - 1) {
sa4 = sa3 = sa2;
}
}
for (x = 0; x < width; x++) {
const uint32_t B1 = sa0[2];
const uint32_t PB = sa1[2];
const uint32_t PE = sa2[2];
const uint32_t PH = sa3[2];
const uint32_t H5 = sa4[2];
const int pprev = 2 - (x > 0);
const uint32_t A1 = sa0[pprev];
const uint32_t PA = sa1[pprev];
const uint32_t PD = sa2[pprev];
const uint32_t PG = sa3[pprev];
const uint32_t G5 = sa4[pprev];
const int pprev2 = pprev - (x > 1);
const uint32_t A0 = sa1[pprev2];
const uint32_t D0 = sa2[pprev2];
const uint32_t G0 = sa3[pprev2];
const int pnext = 3 - (x == width - 1);
const uint32_t C1 = sa0[pnext];
const uint32_t PC = sa1[pnext];
const uint32_t PF = sa2[pnext];
const uint32_t PI = sa3[pnext];
const uint32_t I5 = sa4[pnext];
const int pnext2 = pnext + 1 - (x >= width - 2);
const uint32_t C4 = sa1[pnext2];
const uint32_t F4 = sa2[pnext2];
const uint32_t I4 = sa3[pnext2];
if (n == 2) {
E[0] = E[1] = // 0, 1
E[nl] = E[nl + 1] = PE; // 2, 3
FILT2(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, 0, 1, nl, nl+1);
FILT2(PE, PC, PF, PB, PI, PA, PH, PD, PG, I4, A1, I5, H5, A0, D0, B1, C1, F4, C4, G5, G0, nl, 0, nl+1, 1);
FILT2(PE, PA, PB, PD, PC, PG, PF, PH, PI, C1, G0, C4, F4, G5, H5, D0, A0, B1, A1, I4, I5, nl+1, nl, 1, 0);
FILT2(PE, PG, PD, PH, PA, PI, PB, PF, PC, A0, I5, A1, B1, I4, F4, H5, G5, D0, G0, C1, C4, 1, nl+1, 0, nl);
} else if (n == 3) {
E[0] = E[1] = E[2] = // 0, 1, 2
E[nl] = E[nl+1] = E[nl+2] = // 3, 4, 5
E[nl1] = E[nl1+1] = E[nl1+2] = PE; // 6, 7, 8
FILT3(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, 0, 1, 2, nl, nl+1, nl+2, nl1, nl1+1, nl1+2);
FILT3(PE, PC, PF, PB, PI, PA, PH, PD, PG, I4, A1, I5, H5, A0, D0, B1, C1, F4, C4, G5, G0, nl1, nl, 0, nl1+1, nl+1, 1, nl1+2, nl+2, 2);
FILT3(PE, PA, PB, PD, PC, PG, PF, PH, PI, C1, G0, C4, F4, G5, H5, D0, A0, B1, A1, I4, I5, nl1+2, nl1+1, nl1, nl+2, nl+1, nl, 2, 1, 0);
FILT3(PE, PG, PD, PH, PA, PI, PB, PF, PC, A0, I5, A1, B1, I4, F4, H5, G5, D0, G0, C1, C4, 2, nl+2, nl1+2, 1, nl+1, nl1+1, 0, nl, nl1);
} else if (n == 4) {
E[0] = E[1] = E[2] = E[3] = // 0, 1, 2, 3
E[nl] = E[nl+1] = E[nl+2] = E[nl+3] = // 4, 5, 6, 7
E[nl1] = E[nl1+1] = E[nl1+2] = E[nl1+3] = // 8, 9, 10, 11
E[nl2] = E[nl2+1] = E[nl2+2] = E[nl2+3] = PE; // 12, 13, 14, 15
FILT4(PE, PI, PH, PF, PG, PC, PD, PB, PA, G5, C4, G0, D0, C1, B1, F4, I4, H5, I5, A0, A1, nl2+3, nl2+2, nl1+3, 3, nl+3, nl1+2, nl2+1, nl2, nl1+1, nl+2, 2, 1, nl+1, nl1, nl, 0);
FILT4(PE, PC, PF, PB, PI, PA, PH, PD, PG, I4, A1, I5, H5, A0, D0, B1, C1, F4, C4, G5, G0, 3, nl+3, 2, 0, 1, nl+2, nl1+3, nl2+3, nl1+2, nl+1, nl, nl1, nl1+1, nl2+2, nl2+1, nl2);
FILT4(PE, PA, PB, PD, PC, PG, PF, PH, PI, C1, G0, C4, F4, G5, H5, D0, A0, B1, A1, I4, I5, 0, 1, nl, nl2, nl1, nl+1, 2, 3, nl+2, nl1+1, nl2+1, nl2+2, nl1+2, nl+3, nl1+3, nl2+3);
FILT4(PE, PG, PD, PH, PA, PI, PB, PF, PC, A0, I5, A1, B1, I4, F4, H5, G5, D0, G0, C1, C4, nl2, nl1, nl2+1, nl2+3, nl2+2, nl1+1, nl, 0, nl+1, nl1+2, nl1+3, nl+3, nl+2, 1, 2, 3);
}
sa0 += 1;
sa1 += 1;
sa2 += 1;
sa3 += 1;
sa4 += 1;
E += n;
}
}
}
#include <inttypes.h>
void xbr_filter(uint32_t *data, uint32_t *out, int width, int height, int n);