Unity learning shader notes [82] black and white processing of enhanced single channel color rendering

be based on Unity Study shader note [ Eighty-one ] Simple color adjustment post-processing （ brightness , saturation , Contrast ）

Think of the original form of torture 2 The picture of being beaten with residual blood inside is only reserved for red , Everything else is grey and black . It's similar to this

Got an original picture , As UI Of ScreenSapce-Camera Imaging as a camera , Be careful ScreenSapce-Overlay Of UI Will not participate in screen post-processing .

Because the red part will not turn black , So the idea is that the red part should be the original color , Instead of lack of saturation , The code is as follows

// The stronger the red, the more invariant the gray , The weaker the area, the more use _Saturation To adjust 
float grayDegree = lerp(_Saturation, 1, renderTex.r);
finalColor = lerp(grayColor, finalColor, grayDegree  );					

Set the saturation to 0,
The effect is as follows , Red does obviously remain , But there are other places that retain color .
Because many other colors have r component , We need to be below a certain level r Discard the source of the component r The influence of component

So I made a little change to the code , The idea is to treat the red part below a certain threshold as red 0,
Then treat red as 0 Part of , Use the original red value to adjust the saturation .

float redLerpValue = renderTex.r  - _LeastRed <  0 ? 0 : renderTex.r;
// The stronger the red, the more invariant the gray , The weaker the area, the more use _Saturation To adjust 
float grayDegree = lerp(_Saturation, 1, redLerpValue);
finalColor = lerp(grayColor, finalColor, grayDegree  );

The parameters and effects are as follows ：

It's only red , Just use the original saturation value directly after it is less than a certain value, so that the transition of the film source with the red value above and below the threshold will be very contrary . When I was stuck in difficulties, I suddenly wondered whether these models with special effects were painted with vertex colors when they left the factory , Then the film source makes color reservation according to the value of vertex color , In this way, the parts that should not retain color are strictly limited , But post-processing shader Vertex colors belonging to a single model cannot be read , At most, it is to reconstruct the world coordinates according to the depth map or do some other operations .

Now is to keep the red , If red is lower than a certain value, it turns gray , And the transition is not inconsistent , Think of an effect carefully analyzed before , Unity Grid Programming Notes [ Four ] Press the sphere to compress It's used in pow, Under the same coefficient , He can make the distance between his relatively small decimal and his relatively large decimal , And the larger the coefficient, the larger the drive , But be pow After treatment, the effect is continuous and will not be disconnected, which can simulate the transition

So I made some changes .

	float redLerpValue = pow(renderTex.r, _RedPowerValue) ;
	// The stronger the red, the more invariant the gray , The weaker the area, the more use _Saturation To adjust 
	float grayDegree = lerp(_Saturation, 1, redLerpValue);
	finalColor = lerp(grayColor, finalColor, grayDegree  );

I feel better , But the white part of the sky doesn't turn black and white , The sky is white , White is generally rgb The three channels are relatively high , Then the values of the three channels are equal .

To achieve the effect in the game , It feels like more than just reading r passageway , Yes. rgb Channel comprehensive judgment , Because when it is untreated, the red part will be kept and will not be grayed out , This part must be r The passage is much larger than gb The tunnel , Only considered r Channels are more inclined to single channel color retention , image unity Of inspector When viewing pictures on the panel, you only see the single channel . We use it r The difference between the mean value of the channel and that of other channels is taken as pow The basic value of , This itself is not the red part , The base value will be small and considered to be used only saturation control , Pay attention because the purpose is to highlight the red area , The delimited area of the red area is R The channel is much larger than gb The average of the channels , It's much bigger than , Instead of simply judging the difference between the two , Because of existence (0,1,1) This color , If such a color is also defined as red, it is incorrect .

So I made some changes .

float redChannelValue = renderTex.r;
					float blueChannelValue = renderTex.b;
					float greenChannelValue = renderTex.g;
					float blueGreenAverage = (blueChannelValue + greenChannelValue) / 2;
					
					float redDeltaBlueGreen = abs(redChannelValue - blueGreenAverage) ;
					
					redDeltaBlueGreen = redChannelValue > blueGreenAverage ? redDeltaBlueGreen : 0;
					
					float redGrayLerpValue = pow(redDeltaBlueGreen, _RedPowerValue) ;

					float fullSaturateValue = 1;
					// The stronger the red, the more invariant the gray , The weaker the area, the more use _Saturation To adjust 
					float grayDegree = lerp(_Saturation, fullSaturateValue, redGrayLerpValue);
					finalColor = lerp(grayColor, finalColor, grayDegree  );

The effect is OK , Because it's the difference , Difference as pow It's better than itself r The channel value should be small , As pow When the cardinality of , The degree of attenuation is greater than r The passage should be larger , You can imagine by comparing any left and right points of the function graph , therefore pow The coefficient range of can be reduced . Because before this post-processing, I did some HDR Color +bloom Or special effects , Then I did the black-and-white post-processing , According to the principle of both , There's no interference with each other , The glowing part still glows .

In the game, in order to highlight the desire for blood and survival when being beaten with residual blood, and the contempt for everything else and unclear vision , Red is highlighted , Darken the black and white colors , These changes actually make the dark darker , The light becomes brighter , That is, contrast enhancement , In addition, put Bloom The coefficient of adjustment is larger , Let the red glow more .

Ordinary contrast enhancement is the overall enhancement rather than the obvious enhancement of red contrast , The contrast of the red part here should be specially treated . So I made some changes ：

//contrast Contrast ： First calculate the value with the lowest contrast 
fixed3 avgColor = fixed3(0.5, 0.5, 0.5);


float redContrastLerpValue = pow(redDeltaBlueGreen, 3) ;

float contrastDegree = lerp(_Contrast, _Contrast * 10, redContrastLerpValue);

// according to Contrast The difference between the image with the lowest contrast and the original image 
finalColor = lerp(avgColor, finalColor, contrastDegree);

There may be others when there is residual blood LUT,ToneMapping What's the post-processing of color mapping .

Project address