Nvidia-BiCubic-Interpolation-Sample.cg

/**
 * Taken from https://developer.nvidia.com/gpugems/gpugems2/part-iii-high-quality-rendering/chapter-20-fast-third-order-texture-filtering
 */

float4 bspline_2d_fp(
  float2 coord_source: TEXCOORD0,
  uniform sampler2D tex_source,  // source texture
  uniform sampler1D tex_hg,      // filter offsets and weights
  uniform float2 e_x,            // texel size in x direction
  uniform float2 e_y,            // texel size in y direction
  uniform float2 size_source     // source texture size
) : COLOR {
  // calc filter texture coordinates where [0,1] is a single texel
  // (can be done in vertex program instead)
  float2 coord_hg = coord_source * size_source - float2(0.5f, 0.5f);      // fetch offsets and weights from filter texture
  float3 hg_x = tex1D(tex_hg, coord_hg.x).xyz;
  float3 hg_y = tex1D(tex_hg, coord_hg.y).xyz;      // determine linear sampling coordinates
  float2 coord_source10 = coord_source + hg_x.x * e_x;
  float2 coord_source00 = coord_source - hg_x.y * e_x;
  float2 coord_source11 = coord_source10 + hg_y.x * e_y;
  float2 coord_source01 = coord_source00 + hg_y.x * e_y;
  coord_source10 = coord_source10 - hg_y.y * e_y;
  coord_source00 = coord_source00 - hg_y.y * e_y;      // fetch four linearly interpolated inputs
  float4 tex_source00 = tex2D( tex_source, coord_source00 );
  float4 tex_source10 = tex2D( tex_source, coord_source10 );
  float4 tex_source01 = tex2D( tex_source, coord_source01 );
  float4 tex_source11 = tex2D( tex_source, coord_source11 );      // weight along y direction
  tex_source00 = lerp( tex_source00, tex_source01, hg_y.z );
  tex_source10 = lerp( tex_source10, tex_source11, hg_y.z );      // weight along x direction
  tex_source00 = lerp( tex_source00, tex_source10, hg_x.z );
  
  return tex_src00;
}