koobas.hobune.stream
Shader: texture sample exact pixel value
Is it possible to sample a pixel during shading and getting the exact color values as are stored in the texture image? For example if my first pixel has a red byte value of for example 212 (thats 11010100 in binary), is it possible to get exactly this value during the shading process for a point corresponding to that pixels location?
My problem is, that when I tried the following code and debugged the red value, it was slightly off, as if some rounding or interpolation with nearby pixels occured
CGPROGRAM
#pragma vertex vert
#pragma fragment frag
#include "UnityCG.cginc"
fixed4 _Color;
sampler2D _MainTex;
struct v2f {
float4 pos : SV_POSITION;
float2 uv : TEXCOORD0;
};
float4 _MainTex_ST;
v2f vert(appdata_base v)
{
v2f o;
o.pos = UnityObjectToClipPos(v.vertex);
o.uv = TRANSFORM_TEX(v.texcoord, _MainTex);
return o;
}
fixed4 frag(v2f i) : SV_Target
{
// resolution of the texture is 1024*1024
float texSize = 1024;
// sample the texture but only at positions corresponding to whole pixels
// multiplying the position by the texture width and then flooring it should
// bring it into the [0, 1023] interval with only integer values
// dividing it by the texture width again should return the interval to [0, 1]
float2 myUV = float2(floor(i.uv.x * texSize) / texSize, floor(i.uv.y * texSize) / texSize);
fixed4 texcol = tex2D(_MainTex, myUV);
// I want to get exactly the value that was in the pixel at the sampled location
fixed redPixelValue = floor(texcol.r * 256)
return texcol * _Color;
}
ENDCG
Is there something I am missing? Thank you for any answers!
$$anonymous$$y suggestion would be to use a higher precision format, although there's no guarantee that would work. I've also changed some things just to make it more usable (texSize for any texture, etc.);
CGPROGRA$$anonymous$$
#pragma vertex vert
#pragma fragment frag
#include "UnityCG.cginc"
//fixed4 _Color;
sampler2D _$$anonymous$$ainTex;
struct v2f {
float4 pos : SV_POSITION;
float2 uv : TEXCOORD0;
};
float4 _$$anonymous$$ainTex_ST;
float4 _$$anonymous$$ainTex_TexelSize;
v2f vert(appdata_base v)
{
v2f o;
o.pos = UnityObjectToClipPos(v.vertex);
o.uv = TRANSFOR$$anonymous$$_TEX(v.texcoord, _$$anonymous$$ainTex);
return o;
}
float4 frag(v2f i) : SV_Target
{
float2 texSize = 1/_$$anonymous$$ainTex_TexelSize.xy;
// sample the texture but only at positions corresponding to whole pixels
// multiplying the position by the texture width and then flooring it should
// bring it into the [0, 1023] interval with only integer values
// dividing it by the texture width again should return the interval to [0, 1]
float2 myUV = float2(floor(i.uv.x * texSize.x) / texSize.x, floor(i.uv.y * texSize.y) / texSize.y);
float4 texcol = tex2D(_$$anonymous$$ainTex, myUV);
float redPixelValue = texcol.r;
return float4 (texcol, 0, 0, 1); // * _Color;
}
ENDCG
$$anonymous$$ight help, might not.
I already found the culprit - the texture was being compressed on import.
Though your answer did help me to get a cleaner code. Thank you!
Hi, I'm not sure i understand what you want to do: doesn't using "point" filtering on the texture settings does what you want ? (bilinear or trilinear will always sample an interpolation of multiple texels)
I was already using point filtering, I must be overlooking something else. Thank you for you time, though!
Edit: I finally found it! The texture was compressed on import, I found the setting when I double checked the point filtering.
Setting it to uncompressed ARGB 32 bit did the trick!
Thank you!
oh and you need the center of the pixel, so your values are actually .5 to 1023.5
