CfL_Prediction.glsl

// MIT License

// Copyright (c) 2023 João Chrisóstomo

// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:

// The above copyright notice and this permission notice shall be included in all
// copies or substantial portions of the Software.

// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.

//!PARAM chroma_offset_x
//!TYPE float
0.0

//!PARAM chroma_offset_y
//!TYPE float
0.0

//!HOOK CHROMA
//!BIND LUMA
//!BIND CHROMA
//!SAVE LUMA_LR
//!WIDTH CHROMA.w
//!HEIGHT LUMA.h
//!WHEN CHROMA.w LUMA.w <
//!DESC Chroma From Luma Prediction (Hermite 1st step, Downscaling Luma)

float comp_wd(vec2 v) {
    float x = min(length(v), 1.0);
    return smoothstep(0.0, 1.0, 1.0 - x);
}

vec4 hook() {
    vec2 luma_pos = LUMA_pos;
    luma_pos.x += chroma_offset_x / LUMA_size.x;
    float start  = ceil((luma_pos.x - (1.0 / CHROMA_size.x)) * LUMA_size.x - 0.5);
    float end = floor((luma_pos.x + (1.0 / CHROMA_size.x)) * LUMA_size.x - 0.5);

    float wt = 0.0;
    float luma_sum = 0.0;
    vec2 pos = luma_pos;

    for (float dx = start.x; dx <= end.x; dx++) {
        pos.x = LUMA_pt.x * (dx + 0.5);
        vec2 dist = (pos - luma_pos) * CHROMA_size;
        float wd = comp_wd(dist);
        float luma_pix = LUMA_tex(pos).x;
        luma_sum += wd * luma_pix;
        wt += wd;
    }

    vec4 output_pix = vec4(luma_sum /= wt, 0.0, 0.0, 1.0);
    return clamp(output_pix, 0.0, 1.0);
}

//!HOOK CHROMA
//!BIND LUMA_LR
//!BIND CHROMA
//!BIND LUMA
//!SAVE LUMA_LR
//!WIDTH CHROMA.w
//!HEIGHT CHROMA.h
//!WHEN CHROMA.w LUMA.w <
//!DESC Chroma From Luma Prediction (Hermite 2nd step, Downscaling Luma)

float comp_wd(vec2 v) {
    float x = min(length(v), 1.0);
    return smoothstep(0.0, 1.0, 1.0 - x);
}

vec4 hook() {
    vec2 luma_pos = LUMA_LR_pos;
    luma_pos.y += chroma_offset_y / LUMA_LR_size.y;
    float start  = ceil((luma_pos.y - (1.0 / CHROMA_size.y)) * LUMA_LR_size.y - 0.5);
    float end = floor((luma_pos.y + (1.0 / CHROMA_size.y)) * LUMA_LR_size.y - 0.5);

    float wt = 0.0;
    float luma_sum = 0.0;
    vec2 pos = luma_pos;

    for (float dy = start; dy <= end; dy++) {
        pos.y = LUMA_LR_pt.y * (dy + 0.5);
        vec2 dist = (pos - luma_pos) * CHROMA_size;
        float wd = comp_wd(dist);
        float luma_pix = LUMA_LR_tex(pos).x;
        luma_sum += wd * luma_pix;
        wt += wd;
    }

    vec4 output_pix = vec4(luma_sum /= wt, 0.0, 0.0, 1.0);
    return clamp(output_pix, 0.0, 1.0);
}

//!HOOK CHROMA
//!BIND HOOKED
//!BIND LUMA
//!BIND LUMA_LR
//!WHEN CHROMA.w LUMA.w <
//!WIDTH LUMA.w
//!HEIGHT LUMA.h
//!OFFSET ALIGN
//!DESC Chroma From Luma Prediction (Upscaling Chroma)

#define USE_12_TAP_REGRESSION 1
#define USE_8_TAP_REGRESSIONS 1
#define DEBUG 0

float comp_wd(vec2 v) {
    float d = min(length(v), 2.0);
    float d2 = d * d;
    float d3 = d2 * d;

    if (d < 1.0) {
        return 1.25 * d3 - 2.25 * d2 + 1.0;
    } else {
        return -0.75 * d3 + 3.75 * d2 - 6.0 * d + 3.0;
    }
}

vec4 hook() {
    float ar_strength = 0.8;
    vec2 mix_coeff = vec2(0.8);
    vec2 corr_exponent = vec2(4.0);

    vec4 output_pix = vec4(0.0, 0.0, 0.0, 1.0);
    float luma_zero = LUMA_texOff(0.0).x;

    vec2 pp = HOOKED_pos * HOOKED_size - vec2(0.5);
    vec2 fp = floor(pp);
    pp -= fp;

#ifdef HOOKED_gather
    vec2 quad_idx[4] = {{0.0, 0.0}, {2.0, 0.0}, {0.0, 2.0}, {2.0, 2.0}};

    vec4 luma_quads[4];
    vec4 chroma_quads[4][2];

    for (int i = 0; i < 4; i++) {
        luma_quads[i] = LUMA_LR_gather(vec2((fp + quad_idx[i]) * HOOKED_pt), 0);
        chroma_quads[i][0] = HOOKED_gather(vec2((fp + quad_idx[i]) * HOOKED_pt), 0);
        chroma_quads[i][1] = HOOKED_gather(vec2((fp + quad_idx[i]) * HOOKED_pt), 1);
    }

    vec2 chroma_pixels[16];
    chroma_pixels[0]  = vec2(chroma_quads[0][0].w, chroma_quads[0][1].w);
    chroma_pixels[1]  = vec2(chroma_quads[0][0].z, chroma_quads[0][1].z);
    chroma_pixels[2]  = vec2(chroma_quads[1][0].w, chroma_quads[1][1].w);
    chroma_pixels[3]  = vec2(chroma_quads[1][0].z, chroma_quads[1][1].z);
    chroma_pixels[4]  = vec2(chroma_quads[0][0].x, chroma_quads[0][1].x);
    chroma_pixels[5]  = vec2(chroma_quads[0][0].y, chroma_quads[0][1].y);
    chroma_pixels[6]  = vec2(chroma_quads[1][0].x, chroma_quads[1][1].x);
    chroma_pixels[7]  = vec2(chroma_quads[1][0].y, chroma_quads[1][1].y);
    chroma_pixels[8]  = vec2(chroma_quads[2][0].w, chroma_quads[2][1].w);
    chroma_pixels[9]  = vec2(chroma_quads[2][0].z, chroma_quads[2][1].z);
    chroma_pixels[10] = vec2(chroma_quads[3][0].w, chroma_quads[3][1].w);
    chroma_pixels[11] = vec2(chroma_quads[3][0].z, chroma_quads[3][1].z);
    chroma_pixels[12] = vec2(chroma_quads[2][0].x, chroma_quads[2][1].x);
    chroma_pixels[13] = vec2(chroma_quads[2][0].y, chroma_quads[2][1].y);
    chroma_pixels[14] = vec2(chroma_quads[3][0].x, chroma_quads[3][1].x);
    chroma_pixels[15] = vec2(chroma_quads[3][0].y, chroma_quads[3][1].y);

    float luma_pixels[16];
    luma_pixels[0]  = luma_quads[0].w;
    luma_pixels[1]  = luma_quads[0].z;
    luma_pixels[2]  = luma_quads[1].w;
    luma_pixels[3]  = luma_quads[1].z;
    luma_pixels[4]  = luma_quads[0].x;
    luma_pixels[5]  = luma_quads[0].y;
    luma_pixels[6]  = luma_quads[1].x;
    luma_pixels[7]  = luma_quads[1].y;
    luma_pixels[8]  = luma_quads[2].w;
    luma_pixels[9]  = luma_quads[2].z;
    luma_pixels[10] = luma_quads[3].w;
    luma_pixels[11] = luma_quads[3].z;
    luma_pixels[12] = luma_quads[2].x;
    luma_pixels[13] = luma_quads[2].y;
    luma_pixels[14] = luma_quads[3].x;
    luma_pixels[15] = luma_quads[3].y;
#else
    vec2 pix_idx[16] = {{-0.5,-0.5}, {0.5,-0.5}, {1.5,-0.5}, {2.5,-0.5},
                        {-0.5, 0.5}, {0.5, 0.5}, {1.5, 0.5}, {2.5, 0.5},
                        {-0.5, 1.5}, {0.5, 1.5}, {1.5, 1.5}, {2.5, 1.5},
                        {-0.5, 2.5}, {0.5, 2.5}, {1.5, 2.5}, {2.5, 2.5}};

    float luma_pixels[16];
    vec2 chroma_pixels[16];

    for (int i = 0; i < 16; i++) {
        luma_pixels[i] = LUMA_LR_tex(vec2((fp + pix_idx[i]) * HOOKED_pt)).x;
        chroma_pixels[i] = HOOKED_tex(vec2((fp + pix_idx[i]) * HOOKED_pt)).xy;
    }
#endif

#if (DEBUG == 1)
    vec2 chroma_spatial = vec2(0.5);
    mix_coeff = vec2(1.0);
#else
    float wd[16];
    float wt = 0.0;
    vec2 ct = vec2(0.0);

    vec2 chroma_min = min(min(min(chroma_pixels[5], chroma_pixels[6]), chroma_pixels[9]), chroma_pixels[10]);
    vec2 chroma_max = max(max(max(chroma_pixels[5], chroma_pixels[6]), chroma_pixels[9]), chroma_pixels[10]);

    const int dx[16] = {-1, 0, 1, 2, -1, 0, 1, 2, -1, 0, 1, 2, -1, 0, 1, 2};
    const int dy[16] = {-1, -1, -1, -1, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2};

    for (int i = 0; i < 16; i++) {
        wd[i] = comp_wd(vec2(dx[i], dy[i]) - pp);
        wt += wd[i];
        ct += wd[i] * chroma_pixels[i];
    }

    vec2 chroma_spatial = ct / wt;
    chroma_spatial = clamp(mix(chroma_spatial, clamp(chroma_spatial, chroma_min, chroma_max), ar_strength), 0.0, 1.0);
#endif

#if (USE_12_TAP_REGRESSION == 1 || USE_8_TAP_REGRESSIONS == 1)
    const int i12[12] = {1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14};
    const int i4y[4] = {1, 2, 13, 14};
    const int i4x[4] = {4, 7, 8, 11};
    const int i4[4] = {5, 6, 9, 10};

    float luma_sum_4 = 0.0;
    float luma_sum_4y = 0.0;
    float luma_sum_4x = 0.0;
    vec2 chroma_sum_4 = vec2(0.0);
    vec2 chroma_sum_4y = vec2(0.0);
    vec2 chroma_sum_4x = vec2(0.0);

    for (int i = 0; i < 4; i++) {
        luma_sum_4 += luma_pixels[i4[i]];
        luma_sum_4y += luma_pixels[i4y[i]];
        luma_sum_4x += luma_pixels[i4x[i]];
        chroma_sum_4 += chroma_pixels[i4[i]];
        chroma_sum_4y += chroma_pixels[i4y[i]];
        chroma_sum_4x += chroma_pixels[i4x[i]];
    }

    float luma_avg_12 = (luma_sum_4 + luma_sum_4y + luma_sum_4x) / 12.0;
    float luma_var_12 = 0.0;
    vec2 chroma_avg_12 = (chroma_sum_4 + chroma_sum_4y + chroma_sum_4x) / 12.0;
    vec2 chroma_var_12 = vec2(0.0);
    vec2 luma_chroma_cov_12 = vec2(0.0);

    for (int i = 0; i < 12; i++) {
        luma_var_12 += pow(luma_pixels[i12[i]] - luma_avg_12, 2.0);
        chroma_var_12 += pow(chroma_pixels[i12[i]] - chroma_avg_12, vec2(2.0));
        luma_chroma_cov_12 += (luma_pixels[i12[i]] - luma_avg_12) * (chroma_pixels[i12[i]] - chroma_avg_12);
    }

    vec2 corr = clamp(abs(luma_chroma_cov_12 / max(sqrt(luma_var_12 * chroma_var_12), 1e-6)), 0.0, 1.0);
    mix_coeff = pow(corr, corr_exponent) * mix_coeff;
#endif

#if (USE_12_TAP_REGRESSION == 1)
    vec2 alpha_12 = luma_chroma_cov_12 / max(luma_var_12, 1e-6);
    vec2 beta_12 = chroma_avg_12 - alpha_12 * luma_avg_12;
    vec2 chroma_pred_12 = clamp(alpha_12 * luma_zero + beta_12, 0.0, 1.0);
#endif

#if (USE_8_TAP_REGRESSIONS == 1)
    const int i8y[8] = {1, 2, 5, 6, 9, 10, 13, 14};
    const int i8x[8] = {4, 5, 6, 7, 8, 9, 10, 11};

    float luma_avg_8y = (luma_sum_4 + luma_sum_4y) / 8.0;
    float luma_avg_8x = (luma_sum_4 + luma_sum_4x) / 8.0;
    float luma_var_8y = 0.0;
    float luma_var_8x = 0.0;
    vec2 chroma_avg_8y = (chroma_sum_4 + chroma_sum_4y) / 8.0;
    vec2 chroma_avg_8x = (chroma_sum_4 + chroma_sum_4x) / 8.0;
    vec2 luma_chroma_cov_8y = vec2(0.0);
    vec2 luma_chroma_cov_8x = vec2(0.0);

    for (int i = 0; i < 8; i++) {
        luma_var_8y += pow(luma_pixels[i8y[i]] - luma_avg_8y, 2.0);
        luma_var_8x += pow(luma_pixels[i8x[i]] - luma_avg_8x, 2.0);
        luma_chroma_cov_8y += (luma_pixels[i8y[i]] - luma_avg_8y) * (chroma_pixels[i8y[i]] - chroma_avg_8y);
        luma_chroma_cov_8x += (luma_pixels[i8x[i]] - luma_avg_8x) * (chroma_pixels[i8x[i]] - chroma_avg_8x);
    }

    vec2 alpha_8y = luma_chroma_cov_8y / max(luma_var_8y, 1e-6);
    vec2 alpha_8x = luma_chroma_cov_8x / max(luma_var_8x, 1e-6);
    vec2 beta_8y = chroma_avg_8y - alpha_8y * luma_avg_8y;
    vec2 beta_8x = chroma_avg_8x - alpha_8x * luma_avg_8x;
    vec2 chroma_pred_8y = clamp(alpha_8y * luma_zero + beta_8y, 0.0, 1.0);
    vec2 chroma_pred_8x = clamp(alpha_8x * luma_zero + beta_8x, 0.0, 1.0);
    vec2 chroma_pred_8 = mix(chroma_pred_8y, chroma_pred_8x, 0.5);
#endif

#if (USE_12_TAP_REGRESSION == 1 && USE_8_TAP_REGRESSIONS == 1)
    output_pix.xy = mix(chroma_spatial, mix(chroma_pred_12, chroma_pred_8, 0.5), mix_coeff);
#elif (USE_12_TAP_REGRESSION == 1 && USE_8_TAP_REGRESSIONS == 0)
    output_pix.xy = mix(chroma_spatial, chroma_pred_12, mix_coeff);
#elif (USE_12_TAP_REGRESSION == 0 && USE_8_TAP_REGRESSIONS == 1)
    output_pix.xy = mix(chroma_spatial, chroma_pred_8, mix_coeff);
#else
    output_pix.xy = chroma_spatial;
#endif

    output_pix.xy = clamp(output_pix.xy, 0.0, 1.0);
    return output_pix;
}

//!PARAM distance_coeff
//!TYPE float
//!MINIMUM 0.0
2.0

//!PARAM intensity_coeff
//!TYPE float
//!MINIMUM 0.0
128.0

//!HOOK CHROMA
//!BIND CHROMA
//!BIND LUMA
//!DESC Chroma From Luma Prediction (Smoothing Chroma)

float comp_w(vec2 spatial_distance, float intensity_distance) {
    return max(100.0 * exp(-distance_coeff * pow(length(spatial_distance), 2.0) - intensity_coeff * pow(intensity_distance, 2.0)), 1e-32);
}

vec4 hook() {
    vec4 output_pix = vec4(0.0, 0.0, 0.0, 1.0);
    float luma_zero = LUMA_texOff(0).x;
    float wt = 0.0;
    vec2 ct = vec2(0.0);

    for (int i = -1; i < 2; i++) {
        for (int j = -1; j < 2; j++) {
            vec2 chroma_pixels = CHROMA_texOff(vec2(i, j)).xy;
            float luma_pixels = LUMA_texOff(vec2(i, j)).x;
            float w = comp_w(vec2(i, j), luma_zero - luma_pixels);
            wt += w;
            ct += w * chroma_pixels;
        }
    }

    output_pix.xy = clamp(ct / wt, 0.0, 1.0);
    return output_pix;
}