diff --git a/libobs/data/bicubic_scale.effect b/libobs/data/bicubic_scale.effect index d3bbf085c..b6ba45ac1 100644 --- a/libobs/data/bicubic_scale.effect +++ b/libobs/data/bicubic_scale.effect @@ -33,37 +33,19 @@ struct FragData { VertOut VSDefault(VertData v_in) { VertOut vert_out; - vert_out.uv = v_in.uv; + vert_out.uv = v_in.uv * base_dimension; vert_out.pos = mul(float4(v_in.pos.xyz, 1.0), ViewProj); return vert_out; } -float weight(float x) -{ - float ax = abs(x); - - /* Sharper version. May look better in some cases. B=0, C=0.75 */ - - if (ax < 2.0) { - float six_i = 1.0 / 6.0; - float x_squared = x * x; - if (ax < 1.0) { - return (x_squared * (7.5 * ax + (-13.5))) * six_i + 1.0; - } - - return (x_squared * ((-4.5) * ax + 22.5) + (-36.0) * ax) * six_i + 3.0; - } - - return 0.0; -} - float4 weight4(float x) { + /* Sharper version. May look better in some cases. B=0, C=0.75 */ return float4( - weight(x - 2.0), - weight(x - 1.0), - weight(x), - weight(x + 1.0)); + ((-0.75 * x + 1.5) * x - 0.75) * x, + (1.25 * x - 2.25) * x * x + 1.0, + ((-1.25 * x + 1.5) * x + 0.75) * x, + (0.75 * x - 0.75) * x * x); } float AspectUndistortX(float x, float a) @@ -98,17 +80,17 @@ float4 undistort_line(float4 xpos, float ypos, float4 rowtaps) float4 DrawBicubic(FragData f_in, bool undistort) { - float2 stepxy = base_dimension_i; - float2 pos = f_in.uv + stepxy * 0.5; - float2 f = frac(pos * base_dimension); + float2 pos = f_in.uv; + float2 pos1 = floor(pos - 0.5) + 0.5; + float2 f = pos - pos1; - float4 rowtaps = weight4(1.0 - f.x); - float4 coltaps = weight4(1.0 - f.y); + float4 rowtaps = weight4(f.x); + float4 coltaps = weight4(f.y); - float2 uv0 = (-1.5 - f) * stepxy + pos; - float2 uv1 = uv0 + stepxy; - float2 uv2 = uv1 + stepxy; - float2 uv3 = uv2 + stepxy; + float2 uv1 = pos1 * base_dimension_i; + float2 uv0 = uv1 - base_dimension_i; + float2 uv2 = uv1 + base_dimension_i; + float2 uv3 = uv2 + base_dimension_i; if (undistort) { float4 xpos = float4(uv0.x, uv1.x, uv2.x, uv3.x); @@ -119,11 +101,11 @@ float4 DrawBicubic(FragData f_in, bool undistort) } float u_weight_sum = rowtaps.y + rowtaps.z; - float u_middle_offset = rowtaps.z * stepxy.x / u_weight_sum; + float u_middle_offset = rowtaps.z * base_dimension_i.x / u_weight_sum; float u_middle = uv1.x + u_middle_offset; float v_weight_sum = coltaps.y + coltaps.z; - float v_middle_offset = coltaps.z * stepxy.y / v_weight_sum; + float v_middle_offset = coltaps.z * base_dimension_i.y / v_weight_sum; float v_middle = uv1.y + v_middle_offset; int2 coord_top_left = int2(max(uv0 * base_dimension, 0.5)); diff --git a/libobs/data/lanczos_scale.effect b/libobs/data/lanczos_scale.effect index bc9241438..33a31f394 100644 --- a/libobs/data/lanczos_scale.effect +++ b/libobs/data/lanczos_scale.effect @@ -34,7 +34,7 @@ struct FragData { VertOut VSDefault(VertData v_in) { VertOut vert_out; - vert_out.uv = v_in.uv; + vert_out.uv = v_in.uv * base_dimension; vert_out.pos = mul(float4(v_in.pos.xyz, 1.0), ViewProj); return vert_out; @@ -42,27 +42,26 @@ VertOut VSDefault(VertData v_in) float weight(float x) { - float ax = abs(x); - if (x == 0.0) - return 1.0; - - float radius = 3.0; - if (ax < radius) { - float PIval = 3.14159265358979323846; - float x_pi = x * PIval; - float radius_i = 1.0 / 3.0; - return radius * sin(x_pi) * sin(x_pi * radius_i) / (x_pi * x_pi); - } - - return 0.0; + float x_pi = x * 3.141592654; + return 3.0 * sin(x_pi) * sin(x_pi * (1.0 / 3.0)) / (x_pi * x_pi); } -float3 weight3(float x) +void weight6(float f_neg, out float3 tap012, out float3 tap345) { - return float3( - weight(x * 2.0 - 3.0), - weight(x * 2.0 - 1.0), - weight(x * 2.0 + 1.0)); + tap012 = float3( + weight(f_neg - 2.0), + weight(f_neg - 1.0), + min(1.0, weight(f_neg))); // Replace NaN with 1.0. + tap345 = float3( + weight(f_neg + 1.0), + weight(f_neg + 2.0), + weight(f_neg + 3.0)); + + // Normalize weights + float sum = tap012.x + tap012.y + tap012.z + tap345.x + tap345.y + tap345.z; + float sum_i = 1.0 / sum; + tap012 = tap012 * sum_i; + tap345 = tap345 * sum_i; } float AspectUndistortX(float x, float a) @@ -87,113 +86,101 @@ float4 undistort_pixel(float xpos, float ypos) return image.Sample(textureSampler, undistort_coord(xpos, ypos)); } -float4 undistort_line(float3 xpos012, float3 xpos345, float ypos, float3 rowtap024, - float3 rowtap135) +float4 undistort_line(float3 xpos012, float3 xpos345, float ypos, float3 rowtap012, + float3 rowtap345) { return - undistort_pixel(xpos012.x, ypos) * rowtap024.x + - undistort_pixel(xpos012.y, ypos) * rowtap135.x + - undistort_pixel(xpos012.z, ypos) * rowtap024.y + - undistort_pixel(xpos345.x, ypos) * rowtap135.y + - undistort_pixel(xpos345.y, ypos) * rowtap024.z + - undistort_pixel(xpos345.z, ypos) * rowtap135.z; + undistort_pixel(xpos012.x, ypos) * rowtap012.x + + undistort_pixel(xpos012.y, ypos) * rowtap012.y + + undistort_pixel(xpos012.z, ypos) * rowtap012.z + + undistort_pixel(xpos345.x, ypos) * rowtap345.x + + undistort_pixel(xpos345.y, ypos) * rowtap345.y + + undistort_pixel(xpos345.z, ypos) * rowtap345.z; } float4 DrawLanczos(FragData f_in, bool undistort) { - float2 stepxy = base_dimension_i; - float2 pos = f_in.uv + stepxy * 0.5; - float2 f = frac(pos * base_dimension); + float2 pos = f_in.uv; + float2 pos2 = floor(pos - 0.5) + 0.5; + float2 f_neg = pos2 - pos; - float2 f_rev_half = (-0.5) * f + 0.5; - float3 rowtap024 = weight3(f_rev_half.x); - float3 rowtap135 = weight3(f_rev_half.x + 0.5); - float3 coltap024 = weight3(f_rev_half.y); - float3 coltap135 = weight3(f_rev_half.y + 0.5); + float3 rowtap012, rowtap345; + weight6(f_neg.x, rowtap012, rowtap345); - // Need normalization if divided value near zero - float rowsum = rowtap024.x + rowtap024.y + rowtap024.z + - rowtap135.x + rowtap135.y + rowtap135.z; - float rowsum_i = 1.0 / rowsum; - rowtap024 = rowtap024 * rowsum_i; - rowtap135 = rowtap135 * rowsum_i; - float colsum = coltap024.x + coltap024.y + coltap024.z + - coltap135.x + coltap135.y + coltap135.z; - float colsum_i = 1.0 / colsum; - coltap024 = coltap024 * colsum_i; - coltap135 = coltap135 * colsum_i; + float3 coltap012, coltap345; + weight6(f_neg.y, coltap012, coltap345); - float2 uv0 = (-2.5 - f) * stepxy + pos; - float2 uv1 = uv0 + stepxy; - float2 uv2 = uv1 + stepxy; - float2 uv3 = uv2 + stepxy; - float2 uv4 = uv3 + stepxy; - float2 uv5 = uv4 + stepxy; + float2 uv2 = pos2 * base_dimension_i; + float2 uv1 = uv2 - base_dimension_i; + float2 uv0 = uv1 - base_dimension_i; + float2 uv3 = uv2 + base_dimension_i; + float2 uv4 = uv3 + base_dimension_i; + float2 uv5 = uv4 + base_dimension_i; if (undistort) { float3 xpos012 = float3(uv0.x, uv1.x, uv2.x); float3 xpos345 = float3(uv3.x, uv4.x, uv5.x); - return undistort_line(xpos012, xpos345, uv0.y, rowtap024, rowtap135) * coltap024.x + - undistort_line(xpos012, xpos345, uv1.y, rowtap024, rowtap135) * coltap135.x + - undistort_line(xpos012, xpos345, uv2.y, rowtap024, rowtap135) * coltap024.y + - undistort_line(xpos012, xpos345, uv3.y, rowtap024, rowtap135) * coltap135.y + - undistort_line(xpos012, xpos345, uv4.y, rowtap024, rowtap135) * coltap024.z + - undistort_line(xpos012, xpos345, uv5.y, rowtap024, rowtap135) * coltap135.z; + return undistort_line(xpos012, xpos345, uv0.y, rowtap012, rowtap345) * coltap012.x + + undistort_line(xpos012, xpos345, uv1.y, rowtap012, rowtap345) * coltap012.y + + undistort_line(xpos012, xpos345, uv2.y, rowtap012, rowtap345) * coltap012.z + + undistort_line(xpos012, xpos345, uv3.y, rowtap012, rowtap345) * coltap345.x + + undistort_line(xpos012, xpos345, uv4.y, rowtap012, rowtap345) * coltap345.y + + undistort_line(xpos012, xpos345, uv5.y, rowtap012, rowtap345) * coltap345.z; } - float u_weight_sum = rowtap024.y + rowtap135.y; - float u_middle_offset = rowtap135.y * stepxy.x / u_weight_sum; + float u_weight_sum = rowtap012.z + rowtap345.x; + float u_middle_offset = rowtap345.x * base_dimension_i.x / u_weight_sum; float u_middle = uv2.x + u_middle_offset; - float v_weight_sum = coltap024.y + coltap135.y; - float v_middle_offset = coltap135.y * stepxy.y / v_weight_sum; + float v_weight_sum = coltap012.z + coltap345.x; + float v_middle_offset = coltap345.x * base_dimension_i.y / v_weight_sum; float v_middle = uv2.y + v_middle_offset; float2 coord_limit = base_dimension - 0.5; float2 coord0_f = max(uv0 * base_dimension, 0.5); - float2 coord1_f = coord0_f + 1.0; - float2 coord4_f = min(coord0_f + 4.0, coord_limit); - float2 coord5_f = min(coord0_f + 5.0, coord_limit); + float2 coord1_f = max(uv1 * base_dimension, 0.5); + float2 coord4_f = min(uv4 * base_dimension, coord_limit); + float2 coord5_f = min(uv5 * base_dimension, coord_limit); int2 coord0 = int2(coord0_f); int2 coord1 = int2(coord1_f); int2 coord4 = int2(coord4_f); int2 coord5 = int2(coord5_f); - float4 row0 = image.Load(int3(coord0, 0)) * rowtap024.x; - row0 += image.Load(int3(coord1.x, coord0.y, 0))* rowtap135.x; + float4 row0 = image.Load(int3(coord0, 0)) * rowtap012.x; + row0 += image.Load(int3(coord1.x, coord0.y, 0)) * rowtap012.y; row0 += image.Sample(textureSampler, float2(u_middle, uv0.y)) * u_weight_sum; - row0 += image.Load(int3(coord4.x, coord0.y, 0)) * rowtap024.z; - row0 += image.Load(int3(coord5.x, coord0.y, 0)) * rowtap135.z; - float4 total = row0 * coltap024.x; + row0 += image.Load(int3(coord4.x, coord0.y, 0)) * rowtap345.y; + row0 += image.Load(int3(coord5.x, coord0.y, 0)) * rowtap345.z; + float4 total = row0 * coltap012.x; - float4 row1 = image.Load(int3(coord0.x, coord1.y, 0)) * rowtap024.x; - row1 += image.Load(int3(coord1.x, coord1.y, 0))* rowtap135.x; + float4 row1 = image.Load(int3(coord0.x, coord1.y, 0)) * rowtap012.x; + row1 += image.Load(int3(coord1.x, coord1.y, 0)) * rowtap012.y; row1 += image.Sample(textureSampler, float2(u_middle, uv1.y)) * u_weight_sum; - row1 += image.Load(int3(coord4.x, coord1.y, 0)) * rowtap024.z; - row1 += image.Load(int3(coord5.x, coord1.y, 0)) * rowtap135.z; - total += row1 * coltap135.x; + row1 += image.Load(int3(coord4.x, coord1.y, 0)) * rowtap345.y; + row1 += image.Load(int3(coord5.x, coord1.y, 0)) * rowtap345.z; + total += row1 * coltap012.y; - float4 row23 = image.Sample(textureSampler, float2(uv0.x, v_middle)) * rowtap024.x; - row23 += image.Sample(textureSampler, float2(uv1.x, v_middle))* rowtap135.x; + float4 row23 = image.Sample(textureSampler, float2(uv0.x, v_middle)) * rowtap012.x; + row23 += image.Sample(textureSampler, float2(uv1.x, v_middle)) * rowtap012.y; row23 += image.Sample(textureSampler, float2(u_middle, v_middle)) * u_weight_sum; - row23 += image.Sample(textureSampler, float2(uv4.x, v_middle)) * rowtap024.z; - row23 += image.Sample(textureSampler, float2(uv5.x, v_middle)) * rowtap135.z; + row23 += image.Sample(textureSampler, float2(uv4.x, v_middle)) * rowtap345.y; + row23 += image.Sample(textureSampler, float2(uv5.x, v_middle)) * rowtap345.z; total += row23 * v_weight_sum; - float4 row4 = image.Load(int3(coord0.x, coord4.y, 0)) * rowtap024.x; - row4 += image.Load(int3(coord1.x, coord4.y, 0))* rowtap135.x; + float4 row4 = image.Load(int3(coord0.x, coord4.y, 0)) * rowtap012.x; + row4 += image.Load(int3(coord1.x, coord4.y, 0)) * rowtap012.y; row4 += image.Sample(textureSampler, float2(u_middle, uv4.y)) * u_weight_sum; - row4 += image.Load(int3(coord4.x, coord4.y, 0)) * rowtap024.z; - row4 += image.Load(int3(coord5.x, coord4.y, 0)) * rowtap135.z; - total += row4 * coltap024.z; + row4 += image.Load(int3(coord4.x, coord4.y, 0)) * rowtap345.y; + row4 += image.Load(int3(coord5.x, coord4.y, 0)) * rowtap345.z; + total += row4 * coltap345.y; - float4 row5 = image.Load(int3(coord0.x, coord5.y, 0)) * rowtap024.x; - row5 += image.Load(int3(coord1.x, coord5.y, 0))* rowtap135.x; + float4 row5 = image.Load(int3(coord0.x, coord5.y, 0)) * rowtap012.x; + row5 += image.Load(int3(coord1.x, coord5.y, 0)) * rowtap012.y; row5 += image.Sample(textureSampler, float2(u_middle, uv5.y)) * u_weight_sum; - row5 += image.Load(int3(coord4.x, coord5.y, 0)) * rowtap024.z; - row5 += image.Load(int3(coord5, 0)) * rowtap135.z; - total += row5 * coltap135.z; + row5 += image.Load(int3(coord4.x, coord5.y, 0)) * rowtap345.y; + row5 += image.Load(int3(coord5, 0)) * rowtap345.z; + total += row5 * coltap345.z; return total; }