miloyip / light2d

Light rendering in 2D

light2d

This project illustrates light rendering in 2D with C.

All samples output PNGs with svpng.

License: public domain.

Basic

Use Monte Carol integration and ray marching of signed distance field (SDF) to render a emissive circle.

Source code: basic.c

Uniform sampling (64 samples per pixel):

Uniform sampling with different number of samples per pixel:

Stratified sampling (64 samples per pixel):

Jittered sampling (64 samples per pixel):

Various sampling method side-by-side comparison (64 samples per pixel)::

Constructive Solid Geometry

Source code: csg.c

Use union operation for creating multiple shapes:

Various CSG operations on two circles:

Shapes

Source code: shapes.c

Examples of various shapes defined by SDF:

Reflection

Source code: reflection.c

Test scene with two boxes:

Visualization of SDF gradient, which is approximated by central difference:

Reflection via recursive tracing:

Concave mirror scene generates caustics effect:

Refraction

Source code: refraction.c

Applying Snell's law to compute refraction direction. Total internal reflection is also handled.

Test scenes:

Fresnel Reflectance

Source code: fresnel.c

Applying Fresnel equation to compute reflectance of dielectric medium.

Without Fresnel:

With Fresnel term:

Beer-Lambert

Source code: beerlambert.c beerlambert_color.c

Applying Beer-Lambert law to simulate absorption of light in medimum.

light2d's Issues

关于这方面的知识应该看什么书入门呢？

Makefile无法正常工作

在Manjaro上会报错：
[wurui@computer light2d]$ make
cc basic.c -o basic
/tmp/ccb1Fj9Q.o：在函数‘circleSDF’中：
basic.c:(.text+0x104e)：对‘sqrtf’未定义的引用
/tmp/ccb1Fj9Q.o：在函数‘sample’中：
basic.c:(.text+0x1189)：对‘sinf’未定义的引用
basic.c:(.text+0x119e)：对‘cosf’未定义的引用
/tmp/ccb1Fj9Q.o：在函数‘main’中：
basic.c:(.text+0x1264)：对‘fminf’未定义的引用
collect2: 错误：ld 返回 1
make: *** [<内置>：basic] 错误 1

必须要改成这样，才能正常工作：
TARGETS=basic csg
OUTPUTS=$(addsuffix .png, $(TARGETS))

all: $(TARGETS)
test: $(TARGETS) $(OUTPUTS)

%: %.c
gcc -lm -Wall -O3 -o $@ $&lt;

%.png: %
sh -c "time ./$<"

clean:
rm $(TARGETS) *.png

Lack of proper gamma compression for the output

There's a major but common mistake which makes all the images look wrong. The images are naturally calculated using linear light values but saved as 8-bit sRGB values without any correct conversion from linear values to sRGB (gamma compressed) values. Here's how it should look once that problem is fixed:

This is quite radically different, and you can now see how the beams that cross into the shadows appear as bright as they should. The way to do it is to replace this:
p[0] = p[1] = p[2] = (int)(fminf(sample((float)x / W, (float)y / H) * 255.0f, 255.0f));
with this:
p[0] = p[1] = p[2] = (int)(255. * lsrgb(fminf(sample((float)x / W, (float)y / H), 1.f)));

and add somewhere a function lsrgb that does this:

double lsrgb(double linear)	// converts a [0.0, 1.0] linear value into a [0.0, 1.0] sRGB value
{
	if (linear <= 0.0031308)
		return linear * 12.92;
	else
		return 1.055 * pow(linear, 1.0/2.4) - 0.055;
}

Btw in the image above I also added Gaussian antialiasing which is achieved by adding a Gaussian jitter in the sample() function to x and y independently.