Image Processing - Daytime

An example of the complexity in describing mother nature is modeling and computing the conditions as shown in the below image of a cloudy daytime. Cloudcover is computed in real time using atmospheric modeling. The thicker the atmosphere, the more light reddens and dims because of scattering. This is important in order to create a model of how the sky should look without any clouds.

As the result, the code creates an image of an idealized cloudless sky for a given sun position. Then we subtract this image from the actual camera's image to highlight clouds, and after that, we calculate all pixels marked as clouds to get the percentage of cloudiness.

The code creates a model that leverages scattering and attenuation of light rays in the atmosphere. This list of tasks below is not complete, only to show that measuring cloudiness involves significant computations.

The image below is a processed daytime image, as opposed to  astronomical night. The sun and moon are masked out to eliminate light colored pixels from the sun rays. A transformation matrix has been used to rotate and align pixels with their true position in the sky, to compensate for the position skewed by the fish eye lens.  

Some data values have been optionally superimposed over the image to provide additional relevant information. The constellations, also shown optionally, are superimposed in their correct location for that geographic location and time.

The trees and other obstacles are subtracted from the cloudiness calculation as needed by the particulars of the site. Optional masking of objects, such as an antenna or dome, can be masked by the system administrator.

• Describe the  scattering and attenuation of light rays in the atmosphere
• Describe single ray of light
• Change the behavior of incident light functions depending on Sun's zenith value. Two modes:
  Day Mode - simpler and faster approach for high Sun (zen < 70 deg)
  Night Mode - additional light attenuation for low Sun (zen > 70 deg)
• Calculate the intensity of the light ray scattered in the atmosphere along the way
• Estimated total optical depth for Rayleigh and Mie components
• Consider the anisotropy of the medium (aerosol)
• Calculate Rayleigh phase function
• Calculate Mie phase function (single-lobed Henyey–Greenstein approximate analytic phase function)
• Calculate atmosphere density coefficients
• Calculate the total reduction of light intensity
• Compensate for low altitude flux attenuation (dust etc.)
• Calculate total flux in given point
• Add sky's base glow constant for a night mode
• Render the image of the sky(atmosphere) depending on Sun position and resolution of the model (width, height, sampling)
• Render the sky and the Sun (ray's intensity) pixel-wise
• Return the cloudiness percentage based on the amount of black and white pixels(clouds)