lygia/v1.2.0/filter/sharpen/adaptive)adaptive sharpening. For strenght values between 0.3 <-> 2.0 are a reasonable range
Dependencies:
Use:
sharpen(<SAMPLER_TYPE> texture, <vec2> st, <vec2> renderSize [, float streanght])
#ifndef SHARPENADAPTIVE_TYPE
#ifdef SHARPEN_TYPE
#define SHARPENADAPTIVE_TYPE SHARPEN_TYPE
#else
#define SHARPENADAPTIVE_TYPE vec4
#endif
#endif
#ifndef SHARPENDADAPTIVE_SAMPLER_FNC
#ifdef SHARPEN_SAMPLER_FNC
#define SHARPENDADAPTIVE_SAMPLER_FNC(TEX, UV) SHARPEN_SAMPLER_FNC(TEX, UV)
#else
#define SHARPENDADAPTIVE_SAMPLER_FNC(TEX, UV) SAMPLER_FNC(TEX, UV)
#endif
#endif
#ifndef SHARPENADAPTIVE_ANIME
#define SHARPENADAPTIVE_ANIME false // Only darken edges
#endif
#ifndef FNC_SHARPENAPTIVE
#define FNC_SHARPENAPTIVE
// Soft limit, modified tanh approx
#define SHARPENADAPTIVE_SOFT_LIM(v,s) ( saturate(abs(v/s)*(27.0 + pow(v/s, 2.0))/(27.0 + 9.0*pow(v/s, 2.0)))*s )
// Weighted power mean
#define SHARPENADAPTIVE_WPMEAN(a,b,w) ( pow(w*pow(abs(a), 0.5) + abs(1.0-w)*pow(abs(b), 0.5), 2.0) )
// Get destination pixel values
#define SHARPENADAPTIVE_DXDY(val) ( length(fwidth(val)) ) // edgemul = 2.2
#ifndef SHARPENADAPTIVE_CTRL
// #define SHARPENADAPTIVE_CTRL(RGB) ( dot(RGB*RGB, vec3(0.212655, 0.715158, 0.072187)) )
#define SHARPENADAPTIVE_CTRL(RGB) sharpendAdaptiveControl(RGB)
#endif
float sharpendAdaptiveControl(in vec3 rgb) { return dot(rgb*rgb, vec3(0.212655, 0.715158, 0.072187)); }
float sharpendAdaptiveControl(in vec4 rgba) { return dot(rgba*rgba, vec4(0.212655, 0.715158, 0.072187, 0.0)); }
#define SHARPENADAPTIVE_DIFF(pix) ( abs(blur-c[pix]) )
SHARPENADAPTIVE_TYPE sharpenAdaptive(SAMPLER_TYPE tex, vec2 st, vec2 pixel, float strenght) {
//-------------------------------------------------------------------------------------------------
// Defined values under this row are "optimal" DO NOT CHANGE IF YOU DO NOT KNOW WHAT YOU ARE DOING!
const float curveslope = 0.5 ; // Sharpening curve slope, high edge values
const float L_overshoot = 0.003; // Max light overshoot before compression [>0.001]
const float L_compr_low = 0.167; // Light compression, default (0.167=~6x)
const float D_overshoot = 0.009; // Max dark overshoot before compression [>0.001]
const float D_compr_low = 0.250; // Dark compression, default (0.250=4x)
const float scale_lim = 0.1 ; // Abs max change before compression [>0.01]
const float scale_cs = 0.056; // Compression slope above scale_lim
// [ c22 ]
// [ c24, c9, c23 ]
// [ c21, c1, c2, c3, c18 ]
// [ c19, c10, c4, c0, c5, c11, c16 ]
// [ c20, c6, c7, c8, c17 ]
// [ c15, c12, c14 ]
// [ c13 ]
SHARPENADAPTIVE_TYPE c[25];
c[0] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(0.0, 0.0) * pixel);
c[1] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-1., -1.) * pixel);
c[2] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(0.0, -1.) * pixel);
c[3] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(1.0, -1.) * pixel);
c[4] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-1., 1.0) * pixel);
c[5] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(1.0, 0.0) * pixel);
c[6] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-1., 1.0) * pixel);
c[7] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(0.0, 1.0) * pixel);
c[8] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(1.0, 1.0) * pixel);
c[9] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(0.0, -2.) * pixel);
c[10] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-2., 0.0) * pixel);
c[11] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 2., 0.0) * pixel);
c[12] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 0., 2.0) * pixel);
c[13] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 0., 3.0) * pixel);
c[14] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 1., 2.0) * pixel);
c[15] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-1., 2.0) * pixel);
c[16] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 3., 0.0) * pixel);
c[17] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 2., 1.0) * pixel);
c[18] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 2.,-1.0) * pixel);
c[19] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-3., 0.0) * pixel);
c[20] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-2., 1.0) * pixel);
c[21] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-2.,-1.0) * pixel);
c[22] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 0.,-3.0) * pixel);
c[23] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2( 1.,-2.0) * pixel);
c[24] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + vec2(-1.,-2.0) * pixel);
float e[13];
e[0] = SHARPENADAPTIVE_DXDY(c[0]);
e[1] = SHARPENADAPTIVE_DXDY(c[1]);
e[2] = SHARPENADAPTIVE_DXDY(c[2]);
e[3] = SHARPENADAPTIVE_DXDY(c[3]);
e[4] = SHARPENADAPTIVE_DXDY(c[4]);
e[5] = SHARPENADAPTIVE_DXDY(c[5]);
e[6] = SHARPENADAPTIVE_DXDY(c[6]);
e[7] = SHARPENADAPTIVE_DXDY(c[7]);
e[8] = SHARPENADAPTIVE_DXDY(c[8]);
e[9] = SHARPENADAPTIVE_DXDY(c[9]);
e[10] = SHARPENADAPTIVE_DXDY(c[10]);
e[11] = SHARPENADAPTIVE_DXDY(c[11]);
e[12] = SHARPENADAPTIVE_DXDY(c[12]);
// Blur, gauss 3x3
SHARPENADAPTIVE_TYPE blur = (2.0 * (c[2]+c[4]+c[5]+c[7]) + (c[1]+c[3]+c[6]+c[8]) + 4.0 * c[0]) / 16.0;
// Contrast compression, center = 0.5, scaled to 1/3
float c_comp = saturate(0.266666681 + 0.9*exp2(dot(blur, SHARPENADAPTIVE_TYPE(-7.4/3.0))));
// Edge detection
// Relative matrix weights
// [ 1 ]
// [ 4, 5, 4 ]
// [ 1, 5, 6, 5, 1 ]
// [ 4, 5, 4 ]
// [ 1 ]
float edge = length( 1.38*SHARPENADAPTIVE_DIFF(0)
+ 1.15*(SHARPENADAPTIVE_DIFF(2) + SHARPENADAPTIVE_DIFF(4) + SHARPENADAPTIVE_DIFF(5) + SHARPENADAPTIVE_DIFF(7))
+ 0.92*(SHARPENADAPTIVE_DIFF(1) + SHARPENADAPTIVE_DIFF(3) + SHARPENADAPTIVE_DIFF(6) + SHARPENADAPTIVE_DIFF(8))
+ 0.23*(SHARPENADAPTIVE_DIFF(9) + SHARPENADAPTIVE_DIFF(10) + SHARPENADAPTIVE_DIFF(11) + SHARPENADAPTIVE_DIFF(12)) ) * c_comp;
vec2 cs = vec2(L_compr_low, D_compr_low);
// RGB to luma
float luma[25];
luma[0] = SHARPENADAPTIVE_CTRL(c[0]);
luma[1] = SHARPENADAPTIVE_CTRL(c[1]);
luma[2] = SHARPENADAPTIVE_CTRL(c[2]);
luma[3] = SHARPENADAPTIVE_CTRL(c[3]);
luma[4] = SHARPENADAPTIVE_CTRL(c[4]);
luma[5] = SHARPENADAPTIVE_CTRL(c[5]);
luma[6] = SHARPENADAPTIVE_CTRL(c[6]);
luma[7] = SHARPENADAPTIVE_CTRL(c[7]);
luma[8] = SHARPENADAPTIVE_CTRL(c[8]);
luma[9] = SHARPENADAPTIVE_CTRL(c[9]);
luma[10] = SHARPENADAPTIVE_CTRL(c[10]);
luma[11] = SHARPENADAPTIVE_CTRL(c[11]);
luma[12] = SHARPENADAPTIVE_CTRL(c[12]);
luma[13] = SHARPENADAPTIVE_CTRL(c[13]);
luma[14] = SHARPENADAPTIVE_CTRL(c[14]);
luma[15] = SHARPENADAPTIVE_CTRL(c[15]);
luma[16] = SHARPENADAPTIVE_CTRL(c[16]);
luma[17] = SHARPENADAPTIVE_CTRL(c[17]);
luma[18] = SHARPENADAPTIVE_CTRL(c[18]);
luma[19] = SHARPENADAPTIVE_CTRL(c[19]);
luma[20] = SHARPENADAPTIVE_CTRL(c[20]);
luma[21] = SHARPENADAPTIVE_CTRL(c[21]);
luma[22] = SHARPENADAPTIVE_CTRL(c[22]);
luma[23] = SHARPENADAPTIVE_CTRL(c[23]);
luma[24] = SHARPENADAPTIVE_CTRL(c[24]);
float c0_Y = sqrt(luma[0]);
// Precalculated default squared kernel weights
const vec3 w1 = vec3(0.5, 1.0, 1.41421356237); // 0.25, 1.0, 2.0
const vec3 w2 = vec3(0.86602540378, 1.0, 0.54772255751); // 0.75, 1.0, 0.3
// Transition to a concave kernel if the center edge val is above thr
vec3 dW = pow(mix( w1, w2, saturate(2.4*edge - 0.82)), vec3(2.0));
// Use lower weights for pixels in a more active area relative to center pixel area
// This results in narrower and less visible overshoots around sharp edges
float modif_e0 = 3.0 * e[0] + 0.0090909;
float weights[12];
weights[0] = min(modif_e0/e[1], dW.y);
weights[1] = dW.x;
weights[2] = min(modif_e0/e[3], dW.y);
weights[3] = dW.x;
weights[4] = dW.x;
weights[5] = min(modif_e0/e[6], dW.y);
weights[6] = dW.x;
weights[7] = min(modif_e0/e[8], dW.y);
weights[8] = min(modif_e0/e[9], dW.z);
weights[9] = min(modif_e0/e[10], dW.z);
weights[10] = min(modif_e0/e[11], dW.z);
weights[11] = min(modif_e0/e[12], dW.z);
weights[0] = (max(max((weights[8] + weights[9])/4.0, weights[0]), 0.25) + weights[0])/2.0;
weights[2] = (max(max((weights[8] + weights[10])/4.0, weights[2]), 0.25) + weights[2])/2.0;
weights[5] = (max(max((weights[9] + weights[11])/4.0, weights[5]), 0.25) + weights[5])/2.0;
weights[7] = (max(max((weights[10] + weights[11])/4.0, weights[7]), 0.25) + weights[7])/2.0;
// Calculate the negative part of the laplace kernel and the low threshold weight
float lowthrsum = 0.0;
float weightsum = 0.0;
float neg_laplace = 0.0;
for (int pix = 0; pix < 12; ++pix) {
float lowthr = clamp((29.04*e[pix + 1] - 0.221), 0.01, 1.0);
neg_laplace += luma[pix+1] * weights[pix] * lowthr;
weightsum += weights[pix] * lowthr;
lowthrsum += lowthr / 12.0;
}
neg_laplace = inversesqrt(weightsum / neg_laplace);
// Compute sharpening magnitude function
float sharpen_val = strenght/(strenght*curveslope*pow(edge, 3.5) + 0.625);
// Calculate sharpening diff and scale
float sharpdiff = (c0_Y - neg_laplace)*(lowthrsum*sharpen_val + 0.01);
// Calculate local near min & max, partial sort
float temp = 0.0;
for (int i1 = 0; i1 < 24; i1 += 2) {
temp = luma[i1];
luma[i1] = min(luma[i1], luma[i1+1]);
luma[i1+1] = max(temp, luma[i1+1]);
}
for (int i2 = 24; i2 > 0; i2 -= 2) {
temp = luma[0];
luma[0] = min(luma[0], luma[i2]);
luma[i2] = max(temp, luma[i2]);
temp = luma[24];
luma[24] = max(luma[24], luma[i2-1]);
luma[i2-1] = min(temp, luma[i2-1]);
}
for (int i1 = 1; i1 < 24-1; i1 += 2) {
temp = luma[i1];
luma[i1] = min(luma[i1], luma[i1+1]);
luma[i1+1] = max(temp, luma[i1+1]);
}
for (int i2 = 24-1; i2 > 1; i2 -= 2) {
temp = luma[1];
luma[1] = min(luma[1], luma[i2]);
luma[i2] = max(temp, luma[i2]);
temp = luma[24-1];
luma[24-1] = max(luma[24-1], luma[i2-1]);
luma[i2-1] = min(temp, luma[i2-1]);
}
float nmax = (max(sqrt(luma[23]), c0_Y)*2.0 + sqrt(luma[24]))/3.0;
float nmin = (min(sqrt(luma[1]), c0_Y)*2.0 + sqrt(luma[0]))/3.0;
float min_dist = min(abs(nmax - c0_Y), abs(c0_Y - nmin));
float pos_scale = min_dist + L_overshoot;
float neg_scale = min_dist + D_overshoot;
pos_scale = min(pos_scale, scale_lim*(1.0 - scale_cs) + pos_scale*scale_cs);
neg_scale = min(neg_scale, scale_lim*(1.0 - scale_cs) + neg_scale*scale_cs);
// Soft limited anti-ringing with tanh, SHARPENADAPTIVE_WPMEAN to control compression slope
sharpdiff = (SHARPENADAPTIVE_ANIME ? 0. :
SHARPENADAPTIVE_WPMEAN(max(sharpdiff, 0.0), SHARPENADAPTIVE_SOFT_LIM( max(sharpdiff, 0.0), pos_scale ), cs.x ))
- SHARPENADAPTIVE_WPMEAN(min(sharpdiff, 0.0), SHARPENADAPTIVE_SOFT_LIM( min(sharpdiff, 0.0), neg_scale ), cs.y );
float sharpdiff_lim = saturate(c0_Y + sharpdiff) - c0_Y;
float satmul = (c0_Y + max(sharpdiff_lim*0.9, sharpdiff_lim)*1.03 + 0.03)/(c0_Y + 0.03);
return c0_Y + (sharpdiff_lim*3.0 + sharpdiff)/4.0 + (c[0] - c0_Y)*satmul;
}
SHARPENADAPTIVE_TYPE sharpenAdaptive(SAMPLER_TYPE tex, vec2 st, vec2 pixel) {
return sharpenAdaptive(tex, st, pixel, 1.0);
}
#endif
Dependencies:
Use:
sharpen(<SAMPLER_TYPE> texture, <float2> st, <float2> renderSize [, float streanght])
#ifndef SHARPENADAPTIVE_TYPE
#ifdef SHARPEN_TYPE
#define SHARPENADAPTIVE_TYPE SHARPEN_TYPE
#else
#define SHARPENADAPTIVE_TYPE float4
#endif
#endif
#ifndef SHARPENDADAPTIVE_SAMPLER_FNC
#ifdef SHARPEN_SAMPLER_FNC
#define SHARPENDADAPTIVE_SAMPLER_FNC(TEX, UV) SHARPEN_SAMPLER_FNC(TEX, UV)
#else
#define SHARPENDADAPTIVE_SAMPLER_FNC(TEX, UV) SAMPLER_FNC(TEX, UV)
#endif
#endif
#ifndef SHARPENADAPTIVE_ANIME
#define SHARPENADAPTIVE_ANIME false // Only darken edges
#endif
#ifndef FNC_SHARPENAPTIVE
#define FNC_SHARPENAPTIVE
// Soft limit, modified tanh approx
#define SHARPENADAPTIVE_SOFT_LIM(v,s) ( saturate(abs(v/s)*(27.0 + pow(v/s, 2.0))/(27.0 + 9.0*pow(v/s, 2.0)))*s )
// Weighted power mean
#define SHARPENADAPTIVE_WPMEAN(a,b,w) ( pow(w*pow(abs(a), 0.5) + abs(1.0-w)*pow(abs(b), 0.5), 2.0) )
// Get destination pixel values
#define SHARPENADAPTIVE_DXDY(val) ( length(fwidth(val)) ) // edgemul = 2.2
#ifndef SHARPENADAPTIVE_CTRL
// #define SHARPENADAPTIVE_CTRL(RGB) ( dot(RGB*RGB, float3(0.212655, 0.715158, 0.072187)) )
#define SHARPENADAPTIVE_CTRL(RGB) sharpendAdaptiveControl(RGB)
#endif
float sharpendAdaptiveControl(in float3 rgb) { return dot(rgb*rgb, float3(0.212655, 0.715158, 0.072187)); }
float sharpendAdaptiveControl(in float4 rgba) { return dot(rgba*rgba, float4(0.212655, 0.715158, 0.072187, 0.0)); }
#define SHARPENADAPTIVE_DIFF(pix) ( abs(blur-c[pix]) )
SHARPENADAPTIVE_TYPE sharpenAdaptive(SAMPLER_TYPE tex, float2 st, float2 pixel, float strenght) {
//-------------------------------------------------------------------------------------------------
// Defined values under this row are "optimal" DO NOT CHANGE IF YOU DO NOT KNOW WHAT YOU ARE DOING!
const float curveslope = 0.5 ; // Sharpening curve slope, high edge values
const float L_overshoot = 0.003; // Max light overshoot before compression [>0.001]
const float L_compr_low = 0.167; // Light compression, default (0.167=~6x)
const float D_overshoot = 0.009; // Max dark overshoot before compression [>0.001]
const float D_compr_low = 0.250; // Dark compression, default (0.250=4x)
const float scale_lim = 0.1 ; // Abs max change before compression [>0.01]
const float scale_cs = 0.056; // Compression slope above scale_lim
// [ c22 ]
// [ c24, c9, c23 ]
// [ c21, c1, c2, c3, c18 ]
// [ c19, c10, c4, c0, c5, c11, c16 ]
// [ c20, c6, c7, c8, c17 ]
// [ c15, c12, c14 ]
// [ c13 ]
SHARPENADAPTIVE_TYPE c[25];
c[0] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(0.0, 0.0) * pixel);
c[1] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-1., -1.) * pixel);
c[2] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(0.0, -1.) * pixel);
c[3] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(1.0, -1.) * pixel);
c[4] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-1., 1.0) * pixel);
c[5] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(1.0, 0.0) * pixel);
c[6] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-1., 1.0) * pixel);
c[7] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(0.0, 1.0) * pixel);
c[8] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(1.0, 1.0) * pixel);
c[9] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(0.0, -2.) * pixel);
c[10] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-2., 0.0) * pixel);
c[11] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 2., 0.0) * pixel);
c[12] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 0., 2.0) * pixel);
c[13] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 0., 3.0) * pixel);
c[14] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 1., 2.0) * pixel);
c[15] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-1., 2.0) * pixel);
c[16] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 3., 0.0) * pixel);
c[17] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 2., 1.0) * pixel);
c[18] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 2.,-1.0) * pixel);
c[19] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-3., 0.0) * pixel);
c[20] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-2., 1.0) * pixel);
c[21] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-2.,-1.0) * pixel);
c[22] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 0.,-3.0) * pixel);
c[23] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2( 1.,-2.0) * pixel);
c[24] = SHARPENDADAPTIVE_SAMPLER_FNC(tex, st + float2(-1.,-2.0) * pixel);
float e[13];
e[0] = SHARPENADAPTIVE_DXDY(c[0]);
e[1] = SHARPENADAPTIVE_DXDY(c[1]);
e[2] = SHARPENADAPTIVE_DXDY(c[2]);
e[3] = SHARPENADAPTIVE_DXDY(c[3]);
e[4] = SHARPENADAPTIVE_DXDY(c[4]);
e[5] = SHARPENADAPTIVE_DXDY(c[5]);
e[6] = SHARPENADAPTIVE_DXDY(c[6]);
e[7] = SHARPENADAPTIVE_DXDY(c[7]);
e[8] = SHARPENADAPTIVE_DXDY(c[8]);
e[9] = SHARPENADAPTIVE_DXDY(c[9]);
e[10] = SHARPENADAPTIVE_DXDY(c[10]);
e[11] = SHARPENADAPTIVE_DXDY(c[11]);
e[12] = SHARPENADAPTIVE_DXDY(c[12]);
// Blur, gauss 3x3
SHARPENADAPTIVE_TYPE blur = (2.0 * (c[2]+c[4]+c[5]+c[7]) + (c[1]+c[3]+c[6]+c[8]) + 4.0 * c[0]) / 16.0;
// Contrast compression, center = 0.5, scaled to 1/3
float fu = -7.4/3.0;
float c_comp = saturate(0.266666681 + 0.9*exp2(dot(blur, float4(fu, fu, fu, fu))));
// Edge detection
// Relative matrix weights
// [ 1 ]
// [ 4, 5, 4 ]
// [ 1, 5, 6, 5, 1 ]
// [ 4, 5, 4 ]
// [ 1 ]
float edge = length( 1.38*SHARPENADAPTIVE_DIFF(0)
+ 1.15*(SHARPENADAPTIVE_DIFF(2) + SHARPENADAPTIVE_DIFF(4) + SHARPENADAPTIVE_DIFF(5) + SHARPENADAPTIVE_DIFF(7))
+ 0.92*(SHARPENADAPTIVE_DIFF(1) + SHARPENADAPTIVE_DIFF(3) + SHARPENADAPTIVE_DIFF(6) + SHARPENADAPTIVE_DIFF(8))
+ 0.23*(SHARPENADAPTIVE_DIFF(9) + SHARPENADAPTIVE_DIFF(10) + SHARPENADAPTIVE_DIFF(11) + SHARPENADAPTIVE_DIFF(12)) ) * c_comp;
float2 cs = float2(L_compr_low, D_compr_low);
// RGB to luma
float luma[25];
luma[0] = SHARPENADAPTIVE_CTRL(c[0]);
luma[1] = SHARPENADAPTIVE_CTRL(c[1]);
luma[2] = SHARPENADAPTIVE_CTRL(c[2]);
luma[3] = SHARPENADAPTIVE_CTRL(c[3]);
luma[4] = SHARPENADAPTIVE_CTRL(c[4]);
luma[5] = SHARPENADAPTIVE_CTRL(c[5]);
luma[6] = SHARPENADAPTIVE_CTRL(c[6]);
luma[7] = SHARPENADAPTIVE_CTRL(c[7]);
luma[8] = SHARPENADAPTIVE_CTRL(c[8]);
luma[9] = SHARPENADAPTIVE_CTRL(c[9]);
luma[10] = SHARPENADAPTIVE_CTRL(c[10]);
luma[11] = SHARPENADAPTIVE_CTRL(c[11]);
luma[12] = SHARPENADAPTIVE_CTRL(c[12]);
luma[13] = SHARPENADAPTIVE_CTRL(c[13]);
luma[14] = SHARPENADAPTIVE_CTRL(c[14]);
luma[15] = SHARPENADAPTIVE_CTRL(c[15]);
luma[16] = SHARPENADAPTIVE_CTRL(c[16]);
luma[17] = SHARPENADAPTIVE_CTRL(c[17]);
luma[18] = SHARPENADAPTIVE_CTRL(c[18]);
luma[19] = SHARPENADAPTIVE_CTRL(c[19]);
luma[20] = SHARPENADAPTIVE_CTRL(c[20]);
luma[21] = SHARPENADAPTIVE_CTRL(c[21]);
luma[22] = SHARPENADAPTIVE_CTRL(c[22]);
luma[23] = SHARPENADAPTIVE_CTRL(c[23]);
luma[24] = SHARPENADAPTIVE_CTRL(c[24]);
float c0_Y = sqrt(luma[0]);
// Precalculated default squared kernel weights
const float3 w1 = float3(0.5, 1.0, 1.41421356237); // 0.25, 1.0, 2.0
const float3 w2 = float3(0.86602540378, 1.0, 0.54772255751); // 0.75, 1.0, 0.3
// Transition to a concave kernel if the center edge val is above thr
float3 dW = pow(lerp( w1, w2, saturate(2.4*edge - 0.82)), float3(2.0, 2.0, 2.0));
// Use lower weights for pixels in a more active area relative to center pixel area
// This results in narrower and less visible overshoots around sharp edges
float modif_e0 = 3.0 * e[0] + 0.0090909;
float weights[12];
weights[0] = min(modif_e0/e[1], dW.y);
weights[1] = dW.x;
weights[2] = min(modif_e0/e[3], dW.y);
weights[3] = dW.x;
weights[4] = dW.x;
weights[5] = min(modif_e0/e[6], dW.y);
weights[6] = dW.x;
weights[7] = min(modif_e0/e[8], dW.y);
weights[8] = min(modif_e0/e[9], dW.z);
weights[9] = min(modif_e0/e[10], dW.z);
weights[10] = min(modif_e0/e[11], dW.z);
weights[11] = min(modif_e0/e[12], dW.z);
weights[0] = (max(max((weights[8] + weights[9])/4.0, weights[0]), 0.25) + weights[0])/2.0;
weights[2] = (max(max((weights[8] + weights[10])/4.0, weights[2]), 0.25) + weights[2])/2.0;
weights[5] = (max(max((weights[9] + weights[11])/4.0, weights[5]), 0.25) + weights[5])/2.0;
weights[7] = (max(max((weights[10] + weights[11])/4.0, weights[7]), 0.25) + weights[7])/2.0;
// Calculate the negative part of the laplace kernel and the low threshold weight
float lowthrsum = 0.0;
float weightsum = 0.0;
float neg_laplace = 0.0;
for (int pix = 0; pix < 12; ++pix) {
float lowthr = clamp((29.04*e[pix + 1] - 0.221), 0.01, 1.0);
neg_laplace += luma[pix+1] * weights[pix] * lowthr;
weightsum += weights[pix] * lowthr;
lowthrsum += lowthr / 12.0;
}
neg_laplace = rsqrt(weightsum / neg_laplace);
// Compute sharpening magnitude function
float sharpen_val = strenght/(strenght*curveslope*pow(edge, 3.5) + 0.625);
// Calculate sharpening diff and scale
float sharpdiff = (c0_Y - neg_laplace)*(lowthrsum*sharpen_val + 0.01);
// Calculate local near min & max, partial sort
float temp = 0.0;
for (int i1 = 0; i1 < 24; i1 += 2) {
temp = luma[i1];
luma[i1] = min(luma[i1], luma[i1+1]);
luma[i1+1] = max(temp, luma[i1+1]);
}
for (int i2 = 24; i2 > 0; i2 -= 2) {
temp = luma[0];
luma[0] = min(luma[0], luma[i2]);
luma[i2] = max(temp, luma[i2]);
temp = luma[24];
luma[24] = max(luma[24], luma[i2-1]);
luma[i2-1] = min(temp, luma[i2-1]);
}
for (int i1 = 1; i1 < 24-1; i1 += 2) {
temp = luma[i1];
luma[i1] = min(luma[i1], luma[i1+1]);
luma[i1+1] = max(temp, luma[i1+1]);
}
for (int i2 = 24-1; i2 > 1; i2 -= 2) {
temp = luma[1];
luma[1] = min(luma[1], luma[i2]);
luma[i2] = max(temp, luma[i2]);
temp = luma[24-1];
luma[24-1] = max(luma[24-1], luma[i2-1]);
luma[i2-1] = min(temp, luma[i2-1]);
}
float nmax = (max(sqrt(luma[23]), c0_Y)*2.0 + sqrt(luma[24]))/3.0;
float nmin = (min(sqrt(luma[1]), c0_Y)*2.0 + sqrt(luma[0]))/3.0;
float min_dist = min(abs(nmax - c0_Y), abs(c0_Y - nmin));
float pos_scale = min_dist + L_overshoot;
float neg_scale = min_dist + D_overshoot;
pos_scale = min(pos_scale, scale_lim*(1.0 - scale_cs) + pos_scale*scale_cs);
neg_scale = min(neg_scale, scale_lim*(1.0 - scale_cs) + neg_scale*scale_cs);
// Soft limited anti-ringing with tanh, SHARPENADAPTIVE_WPMEAN to control compression slope
sharpdiff = (SHARPENADAPTIVE_ANIME ? 0. :
SHARPENADAPTIVE_WPMEAN(max(sharpdiff, 0.0), SHARPENADAPTIVE_SOFT_LIM( max(sharpdiff, 0.0), pos_scale ), cs.x ))
- SHARPENADAPTIVE_WPMEAN(min(sharpdiff, 0.0), SHARPENADAPTIVE_SOFT_LIM( min(sharpdiff, 0.0), neg_scale ), cs.y );
float sharpdiff_lim = saturate(c0_Y + sharpdiff) - c0_Y;
float satmul = (c0_Y + max(sharpdiff_lim*0.9, sharpdiff_lim)*1.03 + 0.03)/(c0_Y + 0.03);
return c0_Y + (sharpdiff_lim*3.0 + sharpdiff)/4.0 + (c[0] - c0_Y)*satmul;
}
SHARPENADAPTIVE_TYPE sharpenAdaptive(SAMPLER_TYPE tex, float2 st, float2 pixel) {
return sharpenAdaptive(tex, st, pixel, 1.0);
}
#endif
LYGIA is dual-licensed under the Prosperity License and the Patron License for sponsors and contributors.
Sponsors and contributors are automatically added to the Patron License and they can ignore the any non-commercial rule of the Prosperity Licensed software (please take a look to the exception).
It's also possible to get a permanent comercial license hook to a single and specific version of LYGIA.
Sign up for the news letter bellow, joing the LYGIA's channel on Discord or follow the Github repository