Requires linear-downscaling=no.

SSimDownscaler.glsl

// This library 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 3.0 of the License, or (at your option) any later version.
// 
// This library 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 this library.

//!HOOK POSTKERNEL
//!BIND HOOKED
//!BIND PREKERNEL
//!SAVE L2
//!WIDTH NATIVE_CROPPED.w
//!WHEN NATIVE_CROPPED.h POSTKERNEL.h >
//!COMPONENTS 3
//!DESC SSimDownscaler L2 pass 1

#define axis 1

#define offset      vec2(0,0)

#define MN(B,C,x)   (x < 1.0 ? ((2.-1.5*B-(C))*x + (-3.+2.*B+C))*x*x + (1.-(B)/3.) : (((-(B)/6.-(C))*x + (B+5.*C))*x + (-2.*B-8.*C))*x+((4./3.)*B+4.*C))
#define Kernel(x)   MN(.0, .5, abs(x))
#define taps        2.0

vec4 hook() {
    vec2 base = PREKERNEL_pt * (PREKERNEL_pos * input_size + tex_offset);

    float low  = ceil((PREKERNEL_pos - taps*POSTKERNEL_pt) * input_size - offset + tex_offset - 0.5)[axis];
    float high = floor((PREKERNEL_pos + taps*POSTKERNEL_pt) * input_size - offset + tex_offset - 0.5)[axis];

    float W = 0.0;
    vec4 avg = vec4(0);
    vec2 pos = base;

    for (float k = low; k <= high; k++) {
        pos[axis] = PREKERNEL_pt[axis] * (k - offset[axis] + 0.5);
        float rel = (pos[axis] - base[axis])*POSTKERNEL_size[axis];
        float w = Kernel(rel);

        vec4 tex = textureLod(PREKERNEL_raw, pos, 0.0) * PREKERNEL_mul;
        avg += w * tex * tex;
        W += w;
    }
    avg /= W;

    return avg;
}

//!HOOK POSTKERNEL
//!BIND HOOKED
//!BIND L2
//!SAVE L2
//!WHEN NATIVE_CROPPED.w POSTKERNEL.w >
//!COMPONENTS 3
//!DESC SSimDownscaler L2 pass 2

#define axis 0

#define offset      vec2(0,0)

#define MN(B,C,x)   (x < 1.0 ? ((2.-1.5*B-(C))*x + (-3.+2.*B+C))*x*x + (1.-(B)/3.) : (((-(B)/6.-(C))*x + (B+5.*C))*x + (-2.*B-8.*C))*x+((4./3.)*B+4.*C))
#define Kernel(x)   MN(.0, .5, abs(x))
#define taps        2.0

vec4 hook() {
    float low  = ceil((L2_pos - taps*POSTKERNEL_pt) * L2_size - offset - 0.5)[axis];
    float high = floor((L2_pos + taps*POSTKERNEL_pt) * L2_size - offset - 0.5)[axis];

    float W = 0.0;
    vec4 avg = vec4(0);
    vec2 pos = L2_pos;

    for (float k = low; k <= high; k++) {
        pos[axis] = L2_pt[axis] * (k - offset[axis] + 0.5);
        float rel = (pos[axis] - L2_pos[axis])*POSTKERNEL_size[axis];
        float w = Kernel(rel);

        avg += w * textureLod(L2_raw, pos, 0.0) * L2_mul;
        W += w;
    }
    avg /= W;

    return avg;
}

//!HOOK POSTKERNEL
//!BIND HOOKED
//!BIND L2
//!SAVE MR
//!WHEN NATIVE_CROPPED.h POSTKERNEL.h >
//!COMPONENTS 4
//!DESC SSimDownscaler mean & R

#define oversharp   0.0

#define sigma_nsq   10. / (255.*255.)
#define locality    2.0

#define offset      vec2(0,0)

#define Kernel(x)   pow(1.0 / locality, abs(x))
#define taps        3.0

#define Luma(rgb)   ( dot(rgb, vec3(0.2126, 0.7152, 0.0722)) )

mat3x3 ScaleH(vec2 pos) {
    float low  = ceil(-0.5*taps - offset)[0];
    float high = floor(0.5*taps - offset)[0];

    float W = 0.0;
    mat3x3 avg = mat3x3(0);

    for (float k = low; k <= high; k++) {
        pos[0] = HOOKED_pos[0] + HOOKED_pt[0] * k;
        float rel = k + offset[0];
        float w = Kernel(rel);

        vec3 L = POSTKERNEL_tex(pos).rgb;
        avg += w * mat3x3(L, L*L, L2_tex(pos).rgb);
        W += w;
    }
    avg /= W;

    return avg;
}

vec4 hook() {
    vec2 pos = HOOKED_pos;

    float low  = ceil(-0.5*taps - offset)[1];
    float high = floor(0.5*taps - offset)[1];

    float W = 0.0;
    mat3x3 avg = mat3x3(0);

    for (float k = low; k <= high; k++) {
        pos[1] = HOOKED_pos[1] + HOOKED_pt[1] * k;
        float rel = k + offset[1];
        float w = Kernel(rel);

        avg += w * ScaleH(pos);
        W += w;
    }
    avg /= W;

    float Sl = Luma(max(avg[1] - avg[0] * avg[0], 0.));
    float Sh = Luma(max(avg[2] - avg[0] * avg[0], 0.));
    return vec4(avg[0], mix(sqrt((Sh + sigma_nsq) / (Sl + sigma_nsq)) * (1. + oversharp), clamp(Sh / Sl, 0., 1.), int(Sl > Sh)));
}

//!HOOK POSTKERNEL
//!BIND HOOKED
//!BIND MR
//!WHEN NATIVE_CROPPED.h POSTKERNEL.h >
//!DESC SSimDownscaler final pass

#define locality    2.0

#define offset      vec2(0,0)

#define Kernel(x)   pow(1.0 / locality, abs(x))
#define taps        3.0

#define Gamma(x)    ( pow(x, vec3(1.0/2.0)) )
#define GammaInv(x) ( pow(clamp(x, 0.0, 1.0), vec3(2.0)) )

mat3x3 ScaleH(vec2 pos) {
    float low  = ceil(-0.5*taps - offset)[0];
    float high = floor(0.5*taps - offset)[0];

    float W = 0.0;
    mat3x3 avg = mat3x3(0);

    for (float k = low; k <= high; k++) {
        pos[0] = HOOKED_pos[0] + HOOKED_pt[0] * k;
        float rel = k + offset[0];
        float w = Kernel(rel);

        vec4 MR = MR_tex(pos);
        avg += w * mat3x3(MR.a*MR.rgb, MR.rgb, MR.aaa);
        W += w;
    }
    avg /= W;

    return avg;
}

vec4 hook() {
    vec2 pos = HOOKED_pos;

    float low  = ceil(-0.5*taps - offset)[1];
    float high = floor(0.5*taps - offset)[1];

    float W = 0.0;
    mat3x3 avg = mat3x3(0);

    for (float k = low; k <= high; k++) {
        pos[1] = HOOKED_pos[1] + HOOKED_pt[1] * k;
        float rel = k + offset[1];
        float w = Kernel(rel);

        avg += w * ScaleH(pos);
        W += w;
    }
    avg /= W;
    vec4 L = POSTKERNEL_texOff(0);
    return vec4(avg[1] + avg[2] * L.rgb - avg[0], L.a);
}

I see, thanks for answering. It has been very informative.

Made an account to ask this guestion. This is probably unrelated to SSIMDownscaler but the comments above got me asking, when using catmull_rom only (yes, without ssimdownscaler), is it adviseable to use with linear-downscaling=yes?

It is advisable to use linear-downscaling=yes only with soft scalers (it was in comments above), and catmull_rom isn't a soft scaler.

Ah, i was under the impression that catmull_rom was a softscaler. Thanks for clearing it up.

Been testing vo=gpu-next gpu-api=vlukan lately and I accidentlly found out that SSimDownscaler results are very different than it's on vo=gpu gpu-api=d3d11
The dark line in the anime become thicker(also blurer) and you can see "gliches" around the edge

Can't reproduce.

Hi, @igv You had mentioned here that "lanczos with linear-downscaling=no" would give a sharper result.

But you also mentioned here that this is tuned for mitchell.

When downscaling from 3840x2160 to 2560x1440, which one would yield better results?

lanczos

Thank you for your response

The image in HDR mode becomes oil painting with this one.

Can't confirm.

It doesn't happen anymore, I don't know what happened.

Question from #tomasklaen/uosc#405

My mpv config.

My mpv.conf uses profile=gpu-hq which has linear-downscaling=yes by default. I just tried adding ssimds without switching to linear-downscaling=no and it still seems to work, how come?

@igv: Exactly the same pattern/issue in this shader, explained here https://gist.github.com/igv/a015fc885d5c22e6891820ad89555637?permalink_comment_id=4546672#gistcomment-4546672

@@ -12,8 +12,8 @@
 // License along with this library.

 //!HOOK POSTKERNEL
-//!BIND HOOKED
 //!BIND PREKERNEL
+//!BIND HOOKED
 //!SAVE L2
 //!WIDTH NATIVE_CROPPED.w
 //!WHEN NATIVE_CROPPED.h POSTKERNEL.h >
@@ -53,8 +53,8 @@ vec4 hook() {
 }

 //!HOOK POSTKERNEL
-//!BIND HOOKED
 //!BIND L2
+//!BIND HOOKED
 //!SAVE L2
 //!WHEN NATIVE_CROPPED.w POSTKERNEL.w >
 //!COMPONENTS 3