In order to answer this question, we must first determine why flashlights appear brighter at night than in the daytime. Much of the answer has to do with our eyes.

The pupil is the hole in our eyes that allows light to pass through it; cameras call this hole the aperture. The hole is small, relative to the world, which helps with resolving an image. However, the quantity of light that passes through the hole depends on how large it is. Our iris's can expand and contract to control the size of the pupil. When the pupil is large, more light is allowed to enter the eye; when the pupil is small, less light can enter.

The iris contracts automatically in the presence of bright light, since too much like can damage the retina (the surface of the eye that detects light). However, in the absence of light, the iris slowly relaxes, expanding the pupil. This has the effect of allowing more light to enter the eye, which adds to the apparent brightness of the scene.

So what we need is not to change the overall light levels, but instead apply the equivalent of an iris to the final lighting computations of a scene. That is, we determine the overall illumination at a point, but we then filter out some of this light based on a global setting. In dark scenes, we filter less light, and in bright scenes, we filter more light.

This overall process is called high dynamic range lighting (HDR). It is fairly simple and requires very few additional math computations compared to our current model.

The first step is to end the myth that lights themselves have a global maximum brightness. In all previous examples, our light intensity values lived with the range [0, 1]. Clamping light intensity to this range simply does not mirror reality. In reality, the sun is many orders of magnitude brighter than a flashlight. We must allow for this in our lighting equation.

This also means that our accumulated lighting intensity, the value we originally wrote to the fragment shader output, is no longer on the [0, 1] range. And that poses a problem. We can perform lighting computations with a high dynamic range, but monitors can only display colors on the [0, 1] range. We therefore must map from the HDR to the low dynamic range (LDR).

This part of HDR rendering is called tone mapping. There are many possible tone mapping functions, but we will use one that simulates a flexible aperture. It's quite a simple function, really. First, we pick a maximum intensity value for the scene; intensity values above this will be clamped. Then, we just divide the HDR value by the maximum intensity to get the LDR value.

It is the maximum intensity that we will vary. As the sun goes down, the intensity will go down with it. This will allow the sun to be much brighter in the day than the lights, thus overwhelming their contributions to the scene's illumination. But at night, when the maximum intensity is much lower, the other lights will have an apparently higher brightness.

This is implemented in the HDR Lighting tutorial.

This tutorial controls as the previous one, except that the K key will activate HDR lighting. Pressing L or Shift+L will go back to day or night-time LDR lighting from the last tutorial, respectively.

Figure 12.4. HDR Lighting

The code is quite straightforward. We add a floating-point field to the Light uniform block and the LightBlock struct in C++. Technically, we just steal one of the padding floats, so the size remains the same:

struct LightBlockHDR
{
    glm::vec4 ambientIntensity;
    float lightAttenuation;
    float maxIntensity;
    float padding[2];
    PerLight lights[NUMBER_OF_LIGHTS];
};

We also add a new field to SunlightValue: the maximum light intensity. There is also a new function in the LightManager that computes the HDR-version of the light block: GetLightInformationHDR. Technically, all of this code was already in Light.h/cpp, since these files are shared among all of the tutorials here.