New Super IV Filter

As a proof of concept I tried an alternative method to this video to remove polygon jaggies. It uses a median blur and the nnedi filter on top. It almost completely removes the aliasing, and probably the smearing oil painting look, although a fast blur could be replaced instead for a more natural look. With crt shaders on top it should look pretty natural.

This requires a median blur shader which I couldn’t find, I found only this, so not sure how hard would be to port it to vulkan.

2 Likes

That test image is looking quite nice. I’ll see what I can do about the shader.

EDIT: alright, I added the 3x3 and 5x5 median filtering shaders in the slang/denoisers directory.

3 Likes

Thank you. Looks like a VERY smart median filter which accounts for edge preservation, might be good for PP texture filtering. I used a dumb avisynth median blur “removegrain(22)”, but all code I found is in “c”. I’m not going to ask again, but if someone feels inclined give it a try. Testing modes, seems that mode=22 is the only one that does the trick. Here’s a clip. https://youtu.be/9B4Q2k7XQio

Just in case, I “ported” them to glsl and I took the liberty of optimizing them as they were very unoptimized, specially the 5x5 which made an old GPU of mine crawl and beg for mercy but my version is WAY faster.

median_3x3.glsl:

/*
3x3 Median optimized for GeForce 8800

Morgan McGuire and Kyle Whitson
Williams College

Register allocation tips by Victor Huang Xiaohuang
University of Illinois at Urbana-Champaign

http://graphics.cs.williams.edu


Copyright (c) Morgan McGuire and Williams College, 2006
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:

Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.

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 COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 COPYRIGHT
HOLDER OR CONTRIBUTORS 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.
*/

#if defined(VERTEX)

#if __VERSION__ >= 130
#define COMPAT_VARYING out
#define COMPAT_ATTRIBUTE in
#define COMPAT_TEXTURE texture
#else
#define COMPAT_VARYING varying
#define COMPAT_ATTRIBUTE attribute
#define COMPAT_TEXTURE texture2D
#endif

#ifdef GL_ES
#define COMPAT_PRECISION mediump
#else
#define COMPAT_PRECISION
#endif

COMPAT_ATTRIBUTE vec4 VertexCoord;
COMPAT_ATTRIBUTE vec4 COLOR;
COMPAT_ATTRIBUTE vec4 TexCoord;
COMPAT_VARYING vec4 COL0;
COMPAT_VARYING vec4 TEX0;
COMPAT_VARYING vec4 t1;
COMPAT_VARYING vec4 t2;
COMPAT_VARYING vec4 t3;

vec4 _oPosition1;
uniform mat4 MVPMatrix;
uniform COMPAT_PRECISION int FrameDirection;
uniform COMPAT_PRECISION int FrameCount;
uniform COMPAT_PRECISION vec2 OutputSize;
uniform COMPAT_PRECISION vec2 TextureSize;
uniform COMPAT_PRECISION vec2 InputSize;

void main()
{
   vec4 _oColor;
   vec2 _otexCoord;
   gl_Position = VertexCoord.x * MVPMatrix[0] + VertexCoord.y * MVPMatrix[1] + VertexCoord.z * MVPMatrix[2] + VertexCoord.w * MVPMatrix[3];
   _oPosition1 = gl_Position;
   _oColor = COLOR;
   _otexCoord = TexCoord.xy;
   COL0 = COLOR;
   TEX0.xy = TexCoord.xy;

   vec2 ps = vec2(1.0).xy / TextureSize.xy;
   float dx = ps.x;
   float dy = ps.y;

   t1 = TEX0.xxxy + vec4(-dx, 0, dx, -dy);
   t2 = TEX0.xxxy + vec4(-dx, 0, dx,   0);
   t3 = TEX0.xxxy + vec4(-dx, 0, dx,  dy);
}

#elif defined(FRAGMENT)

#if __VERSION__ >= 130
#define COMPAT_VARYING in
#define COMPAT_TEXTURE texture
out vec4 FragColor;
#else
#define COMPAT_VARYING varying
#define FragColor gl_FragColor
#define COMPAT_TEXTURE texture2D
#endif

#ifdef GL_ES
#ifdef GL_FRAGMENT_PRECISION_HIGH
precision highp float;
#else
precision mediump float;
#endif
#define COMPAT_PRECISION mediump
#else
#define COMPAT_PRECISION
#endif

uniform COMPAT_PRECISION int FrameDirection;
uniform COMPAT_PRECISION int FrameCount;
uniform COMPAT_PRECISION vec2 OutputSize;
uniform COMPAT_PRECISION vec2 TextureSize;
uniform COMPAT_PRECISION vec2 InputSize;
uniform sampler2D Texture;
COMPAT_VARYING vec4 TEX0;
//standard texture sample looks like this: COMPAT_TEXTURE(Texture, TEX0.xy);
COMPAT_VARYING vec4 t1;
COMPAT_VARYING vec4 t2;
COMPAT_VARYING vec4 t3;

#define s2(a, b)                temp = a; a = min(a, b); b = max(temp, b);
#define mn3(a, b, c)            s2(a, b); s2(a, c);
#define mx3(a, b, c)            s2(b, c); s2(a, c);

#define mnmx3(a, b, c)          mx3(a, b, c); s2(a, b);                                   // 3 exchanges
#define mnmx4(a, b, c, d)       s2(a, b); s2(c, d); s2(a, c); s2(b, d);                   // 4 exchanges
#define mnmx5(a, b, c, d, e)    s2(a, b); s2(c, d); mn3(a, c, e); mx3(b, d, e);           // 6 exchanges
#define mnmx6(a, b, c, d, e, f) s2(a, d); s2(b, e); s2(c, f); mn3(a, b, c); mx3(d, e, f); // 7 exchanges

void main()
{
   vec3 v[6];

   v[0] = COMPAT_TEXTURE(Texture, t1.xw).rgb;
   v[1] = COMPAT_TEXTURE(Texture, t1.yw).rgb;
   v[2] = COMPAT_TEXTURE(Texture, t1.zw).rgb;
   v[3] = COMPAT_TEXTURE(Texture, t2.xw).rgb;
   v[4] = COMPAT_TEXTURE(Texture, t2.yw).rgb;
   v[5] = COMPAT_TEXTURE(Texture, t2.zw).rgb;

   // Starting with a subset of size 6, remove the min and max each time
   vec3 temp;
   mnmx6(v[0], v[1], v[2], v[3], v[4], v[5]);

   v[5] = COMPAT_TEXTURE(Texture, t3.xw).rgb;

   mnmx5(v[1], v[2], v[3], v[4], v[5]);

   v[5] = COMPAT_TEXTURE(Texture, t3.yw).rgb;

   mnmx4(v[2], v[3], v[4], v[5]);

   v[5] = COMPAT_TEXTURE(Texture, t3.zw).rgb;

   mnmx3(v[3], v[4], v[5]);
   FragColor = vec4(v[4], 1.0);
}
#endif

median_5x5.glsl:

/*
5x5 Median

GLSL 1.0
Morgan McGuire and Kyle Whitson, 2006
Williams College
http://graphics.cs.williams.edu

Copyright (c) Morgan McGuire and Williams College, 2006
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:

Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.

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 COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 COPYRIGHT
HOLDER OR CONTRIBUTORS 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.
*/

#if defined(VERTEX)

#if __VERSION__ >= 130
#define COMPAT_VARYING out
#define COMPAT_ATTRIBUTE in
#define COMPAT_TEXTURE texture
#else
#define COMPAT_VARYING varying
#define COMPAT_ATTRIBUTE attribute
#define COMPAT_TEXTURE texture2D
#endif

#ifdef GL_ES
#define COMPAT_PRECISION mediump
#else
#define COMPAT_PRECISION
#endif

COMPAT_ATTRIBUTE vec4 VertexCoord;
COMPAT_ATTRIBUTE vec4 COLOR;
COMPAT_ATTRIBUTE vec4 TexCoord;
COMPAT_VARYING vec4 COL0;
COMPAT_VARYING vec4 TEX0;
COMPAT_VARYING vec4 t01;
COMPAT_VARYING vec4 t02;
COMPAT_VARYING vec4 t03;
COMPAT_VARYING vec4 t04;
COMPAT_VARYING vec4 t05;
COMPAT_VARYING vec4 t06;
COMPAT_VARYING vec4 t07;
COMPAT_VARYING vec4 t08;
COMPAT_VARYING vec4 t09;
COMPAT_VARYING vec4 t10;
COMPAT_VARYING vec4 t11;
COMPAT_VARYING vec4 t12;

vec4 _oPosition1;
uniform mat4 MVPMatrix;
uniform COMPAT_PRECISION int FrameDirection;
uniform COMPAT_PRECISION int FrameCount;
uniform COMPAT_PRECISION vec2 OutputSize;
uniform COMPAT_PRECISION vec2 TextureSize;
uniform COMPAT_PRECISION vec2 InputSize;

void main()
{
   vec4 _oColor;
   vec2 _otexCoord;
   gl_Position = VertexCoord.x * MVPMatrix[0] + VertexCoord.y * MVPMatrix[1] + VertexCoord.z * MVPMatrix[2] + VertexCoord.w * MVPMatrix[3];
   _oPosition1 = gl_Position;
   _oColor = COLOR;
   _otexCoord = TexCoord.xy;
   COL0 = COLOR;
   TEX0.xy = TexCoord.xy;

   vec2 ps = vec2(1.0).xy / TextureSize.xy;
   float dx1 = ps.x;
   float dx2 = ps.x + ps.x;
   float dy1 = ps.y;
   float dy2 = ps.y + ps.y;

   t01 = TEX0.xyxy + vec4(-dx2, -dy2, -dx1, -dy2);
   t02 = TEX0.xyxy + vec4(   0, -dy2,  dx1, -dy2);
   t03 = TEX0.xyxy + vec4( dx2, -dy2, -dx2, -dy1);
   t04 = TEX0.xyxy + vec4(-dx1, -dy1,    0, -dy1);
   t05 = TEX0.xyxy + vec4( dx1, -dy1,  dx2, -dy1);
   t06 = TEX0.xyxy + vec4(-dx2,    0, -dx1,    0);

   t07 = TEX0.xyxy + vec4( dx1,    0,  dx2,    0);
   t08 = TEX0.xyxy + vec4(-dx2,  dy1, -dx1,  dy1);
   t09 = TEX0.xyxy + vec4(   0,  dy1,  dx1,  dy1);
   t10 = TEX0.xyxy + vec4( dx2,  dy1, -dx2,  dy2);
   t11 = TEX0.xyxy + vec4(-dx1,  dy2,    0,  dy2);
   t12 = TEX0.xyxy + vec4( dx1,  dy2,  dx2,  dy2);
}

#elif defined(FRAGMENT)

#if __VERSION__ >= 130
#define COMPAT_VARYING in
#define COMPAT_TEXTURE texture
out vec4 FragColor;
#else
#define COMPAT_VARYING varying
#define FragColor gl_FragColor
#define COMPAT_TEXTURE texture2D
#endif

#ifdef GL_ES
#ifdef GL_FRAGMENT_PRECISION_HIGH
precision highp float;
#else
precision mediump float;
#endif
#define COMPAT_PRECISION mediump
#else
#define COMPAT_PRECISION
#endif

uniform COMPAT_PRECISION int FrameDirection;
uniform COMPAT_PRECISION int FrameCount;
uniform COMPAT_PRECISION vec2 OutputSize;
uniform COMPAT_PRECISION vec2 TextureSize;
uniform COMPAT_PRECISION vec2 InputSize;
uniform sampler2D Texture;
COMPAT_VARYING vec4 TEX0;
COMPAT_VARYING vec4 t01;
COMPAT_VARYING vec4 t02;
COMPAT_VARYING vec4 t03;
COMPAT_VARYING vec4 t04;
COMPAT_VARYING vec4 t05;
COMPAT_VARYING vec4 t06;
COMPAT_VARYING vec4 t07;
COMPAT_VARYING vec4 t08;
COMPAT_VARYING vec4 t09;
COMPAT_VARYING vec4 t10;
COMPAT_VARYING vec4 t11;
COMPAT_VARYING vec4 t12;

//standard texture sample looks like this: COMPAT_COMPAT_TEXTURE(Texture, TEX0.xy);
#define SourceSize      vec4(TextureSize, 1.0 / TextureSize)

#define s2(a, b)                                temp = a; a = min(a, b); b = max(temp, b);
#define t2(a, b)                                s2(v[a], v[b]);
#define t24(a, b, c, d, e, f, g, h)             t2(a, b); t2(c, d); t2(e, f); t2(g, h);
#define t25(a, b, c, d, e, f, g, h, i, j)       t24(a, b, c, d, e, f, g, h); t2(i, j);

void main()
{
   vec3 v[25];
   /*
   // Add the pixels which make up our window to the pixel array.
   for(int dX = -2; dX <= 2; ++dX) {
       for(int dY = -2; dY <= 2; ++dY) {
           vec2 offset = vec2(float(dX), float(dY));

           // If a pixel in the window is located at (x+dX, y+dY), put it at index (dX + R)(2R + 1) + (dY + R) of the
           // pixel array. This will fill the pixel array, with the top left pixel of the window at pixel[0] and the
           // bottom right pixel of the window at pixel[N-1].
           v[(dX + 2) * 5 + (dY + 2)] = COMPAT_TEXTURE(Texture, TEX0.xy + offset * SourceSize.zw).rgb;
       }
   }
   */
   v[0]  = COMPAT_TEXTURE(Texture, t01.xy).rgb;
   v[5]  = COMPAT_TEXTURE(Texture, t01.zw).rgb;
   v[10] = COMPAT_TEXTURE(Texture, t02.xy).rgb;
   v[15] = COMPAT_TEXTURE(Texture, t02.zw).rgb;
   v[20] = COMPAT_TEXTURE(Texture, t03.xy).rgb;
   v[1]  = COMPAT_TEXTURE(Texture, t03.zw).rgb;
   v[6]  = COMPAT_TEXTURE(Texture, t04.xy).rgb;
   v[11] = COMPAT_TEXTURE(Texture, t04.zw).rgb;
   v[16] = COMPAT_TEXTURE(Texture, t05.xy).rgb;
   v[21] = COMPAT_TEXTURE(Texture, t05.zw).rgb;
   v[2]  = COMPAT_TEXTURE(Texture, t06.xy).rgb;
   v[7]  = COMPAT_TEXTURE(Texture, t06.zw).rgb;
   v[12] = COMPAT_TEXTURE(Texture, TEX0.xy).rgb;
   v[17] = COMPAT_TEXTURE(Texture, t07.xy).rgb;
   v[22] = COMPAT_TEXTURE(Texture, t07.zw).rgb;
   v[3]  = COMPAT_TEXTURE(Texture, t08.xy).rgb;
   v[8]  = COMPAT_TEXTURE(Texture, t08.zw).rgb;
   v[13] = COMPAT_TEXTURE(Texture, t09.xy).rgb;
   v[18] = COMPAT_TEXTURE(Texture, t09.zw).rgb;
   v[23] = COMPAT_TEXTURE(Texture, t10.xy).rgb;
   v[4]  = COMPAT_TEXTURE(Texture, t10.zw).rgb;
   v[9]  = COMPAT_TEXTURE(Texture, t11.xy).rgb;
   v[14] = COMPAT_TEXTURE(Texture, t11.zw).rgb;
   v[19] = COMPAT_TEXTURE(Texture, t12.xy).rgb;
   v[24] = COMPAT_TEXTURE(Texture, t12.zw).rgb;

   vec3 temp;

   t25(0, 1,         3, 4,       2, 4,       2, 3,       6, 7);
   t25(5, 7,         5, 6,       9, 7,       1, 7,       1, 4);
   t25(12, 13,       11, 13,     11, 12,     15, 16,     14, 16);
   t25(14, 15,       18, 19,     17, 19,     17, 18,     21, 22);
   t25(20, 22,       20, 21,     23, 24,     2, 5,       3, 6);
   t25(0, 6,         0, 3,       4, 7,       1, 7,       1, 4);
   t25(11, 14,       8, 14,      8, 11,      12, 15,     9, 15);
   t25(9, 12,        13, 16,     10, 16,     10, 13,     20, 23);
   t25(17, 23,       17, 20,     21, 24,     18, 24,     18, 21);
   t25(19, 22,       8, 17,      9, 18,      0, 18,      0, 9);
   t25(10, 19,       1, 19,      1, 10,      11, 20,     2, 20);
   t25(2, 11,        12, 21,     3, 21,      3, 12,      13, 22);
   t25(4, 22,        4, 13,      14, 23,     5, 23,      5, 14);
   t25(15, 24,       6, 24,      6, 15,      7, 16,      7, 19);
   t25(3, 11,        5, 17,      11, 17,     9, 17,      4, 10);
   t25(6, 12,        7, 14,      4, 6,       4, 7,       12, 14);
   t25(10, 14,       6, 7,       10, 12,     6, 10,      6, 17);
   t25(12, 17,       7, 17,      7, 10,      12, 18,     7, 12);
   t24(10, 18,       12, 20,     10, 20,     10, 12);

   FragColor = vec4(v[12], 1.0);
}
#endif

EDIT: Sorry, I forgot the copyright notice in one of the shaders :sweat:

2 Likes

Nice, thanks for doing that! I didn’t bother optimizing the slang ones since I got 3k fps with the 3x3 and 2k on the 5x5 on a 640x480 image, though moving those calculations to the vertex is what I had planned to do if needed.

1 Like

Also removing accessing packed arrays with a nonconstant array index, dependent texture reads and manually unrolling loops helped. By the way, the original 5x5 was slower than slow-bilateral.glsl :exploding_head:, go figure

dang, really?? On my GPU, slow bilateral is 1/4 the speed of unoptimized 5x5 (~500 fps vs ~2000), and applying the optimizations to the slang version didn’t net any appreciable speedups (~2100 fps vs ~2000). I went ahead and pushed the changes up to the repo anyway, though, in case my experience is an exception for whatever reason (and saving cycles is always a good thing) :slight_smile:

1 Like

Giving a stab at this, I think I’m failing at the min max procedure, also not sure about the sampling order; a1 top-left, a8 bottom-right?

RG_FORCEINLINE Byte rg_mode22_cpp(const Byte* pSrc, int srcPitch) {
    LOAD_SQUARE_CPP(pSrc, srcPitch);

    Byte l1 = (a1 + a8 + 1) / 2;
    Byte l2 = (a2 + a7 + 1) / 2;
    Byte l3 = (a3 + a6 + 1) / 2;
    Byte l4 = (a4 + a5 + 1) / 2;

    auto ma = std::max(std::max(std::max(l1, l2), l3), l4);
    auto mi = std::min(std::min(std::min(l1, l2), l3), l4);

    return clip(c, mi, ma);
}

Based on median3x3.slang:

layout(push_constant) uniform Push
{
	vec4 SourceSize;
	vec4 OriginalSize;
	vec4 OutputSize;
	uint FrameCount;
} params;

layout(std140, set = 0, binding = 0) uniform UBO
{
	mat4 MVP;
} global;

#define s1(a, b)             temp = a; min(temp, b);
#define s2(a, b)             temp = a; max(temp, b);
#define mn3(a, b, c)     s1(a, b); s1(a, c);
#define mx3(a, b, c)     s2(b, c); s2(a, c);

#pragma stage vertex
layout(location = 0) in vec4 Position;
layout(location = 1) in vec2 TexCoord;
layout(location = 0) out vec2 vTexCoord;
layout(location = 1) out vec4 t1;
layout(location = 2) out vec4 t2;
layout(location = 3) out vec4 t3;

void main()
{
   gl_Position = global.MVP * Position;
   vTexCoord = TexCoord;
   float dx = params.SourceSize.z;
   float dy = params.SourceSize.w;

   t1 = vTexCoord.xxxy + vec4(-dx, 0, dx, -dy);
   t2 = vTexCoord.xxxy + vec4(-dx, 0, dx,   0);
   t3 = vTexCoord.xxxy + vec4(-dx, 0, dx,  dy);
}

#pragma stage fragment
layout(location = 0) in vec2 vTexCoord;
layout(location = 1) in vec4 t1;
layout(location = 2) in vec4 t2;
layout(location = 3) in vec4 t3;
layout(location = 0) out vec4 FragColor;
layout(set = 0, binding = 2) uniform sampler2D Source;

void main()
{

  vec4 color = texture(Source, vTexCoord);
  vec3 temp;

  v[0] = texture(Source, (t1.xw + t3.zw + 1.) / 2.).rgb;
  v[1] = texture(Source, (t1.yw + t3.yw + 1.) / 2.).rgb;
  v[2] = texture(Source, (t1.zw + t3.xw + 1.) / 2.).rgb;
  v[3] = texture(Source, (t2.xw + t2.zw + 1.) / 2.).rgb;
  float mn = mn3(v[0], v[1], v[2], v[3]);
  float mx = mx3(v[0], v[1], v[2], v[3]);

  //    auto ma = max(max(max(v[0], v[1]), v[2]), v[3]);
  //    auto mi = min(min(min(v[0], v[1]), v[2]), v[3]);

  //    return clip(c, mi, ma);


  FragColor = vec4(clamp(color.rgb, mn, mx), 1.0);
}

Finally managed to port the filter, just wanted to share the excitement. I will now append the nnedi pass to smooth out the result and pass it over an edge mask to avoid texture smearing.

4 Likes

Nice work, looking forward to the finished product.

Sort of related to this; have there been other attempts to replicate a console’s internal filtering? I know the SNES and Genesis PPUs both did some kind of filtering to the image which really helped graphically complex games look better (Squaresoft games come to mind).

1 Like

I find this filter specially great for those that want HD-ish graphics while using the new parallel. But I think nnedi3 needs an antialiaser beforehand to remove some bumps, the current ones don’t work nice, I need something like SangNom2.

With something like that you can have nice graphics and FPS on low power devices like smartphones because you can in turn disable the core’s VI filtering option. I also think that expanding the filter to a 3D convolution (temporal) might smooth out some pixel flicker.

It’s not just for crt-less shaders, I implemented it into my N64 CRT preset to smooth out the jaggies a little bit, it’s not too evident with screenshots (with all the NTSC passes and such). Also I recommend to use this with angrylion’s “Dedither” filter only otherwise you can get some banding.

2 Likes

I experimented a bit with scalefx, combined it with removegrain22, it enhances the image although I’m not a fan of this kind of smearing so I look forward to port SangNom2 over slang.

All shots with Video bilinear disabled:

removegrain22 + advanced-AA + nnedi3

scalefx+rAA+aa only. Smooths out some diagonals.

So I tried to combine both rg22 and scalefx+rAA+aa

By the way, I noticed that there’s no “Dedither” option in parallel without AA, could it be possible to implement that to the core?

2 Likes

WTH is up with Mario’s eyes, lol. I never noticed how weird they looked.

Looking very nice! Time to start getting into the N64 again.

1 Like

@Nesguy Shhhh, we don’t talk Mario’s opium addiction here, :joy:.

It looks great @Dogway, I’ll mess with this some when you’re further along development.

2 Likes

It’s good as it is in my opinion, I’m already using the preset block after grade and before the ntsc passes, it helps smooth upfront jaggies (those in the back are an impossible mess). The addition of sangnom2 is really nitpicking and it won’t make that much of a difference.

Now that I learned how to do convolutions, I might pass it through a Sobel edge mask so textures don’t smear too much. All I say is that I won’t be working on shaders as much as before lol.

2 Likes

Ohhh, we’ll I’ll mess with it when I get some free time then.

Do you have any preset chains for this with a crt shader added? (I’d assume I’m just run this at the start of the chain.)

EDIT: Question is moot, need to read better lol.

1 Like

My shader is too long, I think I broke a new record stacking 21 shaders lol.

Here is in pastebin. I forgot to set LUT Size to 16, so fix that before trying. I’m also using the crt_royale_braun version with BW mask. And remember to set video_smooth = false.

The thing is subtle, you will notice in movement that there’s somewhat less pixel flicker.

EDIT: And set space_out to 0, -1 is my personal display matrix.

2 Likes

Yeah, some of us are getting close to pushing RA’s shader system to the limit, lol. (I think I’ve almost hit the setting limit with my chain, I’m going to end up with HyperSpaceMaddness’s issue soon, lol.) I did get the shader count down some thanks to you though, lol.

1 Like