Version 0.1.0



htrdr - image renderer of cloudy atmospheres


htrdr [option]... -a atmosphere -m mie


htrdr is an image renderer in the visible part of the spectrum, for scenes composed of an atmospheric gas mixture, clouds, and a ground. It uses spectral data that should be provided for the pressure and temperature atmospheric vertical profile [1] (-a atmosphere), the liquid water content in suspension within the clouds stored in a htcp(5) file (-c clouds), and the optical properties of water droplets that have been obtained from a Mie code and formatted according to the htmie(5) file format (-m mie). The user also has to provide: the characteristics of the simulated camera (-C camera), the sensor definition (-i image), and the position of the sun (-D azimuth,elevation). It is also possible to provide an OBJ [2] representing the ground geometry with a diffuse reflectivity (-g ground). Both, the clouds and the ground, can be infinitely repeated along the X and Y axis by setting the -r and the -R options, respectively.

htrdr evaluates the intensity incoming on each pixel of the sensor array. The underlying algorithm is based on a Monte-Carlo method: it consists in simulating a given number of optical paths originating from the camera, directed into the atmosphere, taking into account light absorption and scattering phenomena. The computation is performed over the whole visible part of the spectrum, for the three components of the CIE 1931 XYZ colorimetric space that are subsequently recombined in order to obtain the final color for each pixel, and finally the whole image of the scene as seen from the required observation position.

In htrdr the spatial unit 1.0 corresponds to one meter. The estimated radiance of each pixel component is given in The pixels are written into the output file or to the standard output whether the -o option is defined or not, respectively. The output image is a list of raw ASCII data formatted with respect to the htrdr-image(5) file format. Since htrdr relies on the Monte-Carlo method, the estimated radiance of a pixel component is provided with its numerical accuracy.

During simulation, htrdr dynamically loads/unloads cloud properties to handle clouds whose data do not feat in main memory. htrdr also supports shared memory parallelism and relies on the Message Passing Interface specification [4] to parallelise its computations in a distribute memory environment; it can thus be run either directly or through a MPI process launcher like mpirun(1).


-a atmosphere

Path toward the file containing the gas optical properties of the atmosphere. Data must be formatted according to the fileformat described in [1].

-c clouds

Submit a htcp(5) file describing the properties of the clouds. If not defined, only the atmopshere properties submitted through the -a option are taken into account.

-C <camera-parameter:...>

Define the rendering point of view. Available camera parameters are:


Vertical field of view of the camera in [30, 120] degrees. By default angle is set to 70 degrees.


Position of the camera. By default it is set to {0,0,0}.


Position targeted by the camera. By default it is set to {0,1,0}.


Up vector of the camera. By default it is set to {0,0,1}.

-D <azimuth,elevation>

Direction toward the sun center. The direction is defined by two angles in degrees: the azimuth angle in [0, 360[ and the elevation angle in [0, 90]. Following the right-handed convention, the azimuthal rotation is counter-clockwise, with 0 degree on the X axis. The elevation starts from 0 degree for a direction in the XY plane, up to 90 degrees at zenith. Thus -D0,0 -D90,0 -D180,0 and -D270,0 will produce solar vectors {+1,0,0} {0,+1,0} {-1,0,0} and {0,-1,0} respectively, while -Dazimuth,90 will produce {0,0,+1} regardless of azimuth value.


Write in output the space partitionning data structures used to speed up the radiative transfer computations in the clouds. The written data are octrees saved in the VTK file format [3]. Each octree node stores the minimum and the maximum of the extinction coefficients of the cloud cells overlapped by the octree node. In the output file, each octree is separated of the previous one by a line with three minus characters, i.e. ---.

-e reflectivity

Reflectivity of the ground geometry in [0, 1]. By default it is set to 0.5. This parameter is fixed for the while visible range.


Force overwrite of the output file.

-g ground

Path toward an OBJ file [2] representing the ground geometry.


Pre-compute or use cached grids of the cloud properties built from the clouds, the atmosphere and the mie files. If the corresponding grids were generated in a previous run, reuse them as far as it is possible, i.e. if the clouds, the atmosphere and the mie files were not updated. The cached data are written in a hidden directory named .htrdr located in the directory where htrdr is run. On platforms with an efficient hard-drive and plenty of random access memory, this cache mechanism can significantly speed-up the pre-computation step on clouds data. Note that this option is incompatible with a MPI execution and is thus forced to off if htrdr is run through a process launcher.


List short help and exit.

-i <image-parameter:...>

Define the image to render. Available image parameters are:


Definition of the image. By default the image definition is 320x240.


Number of samples per pixel and per component. i.e. the estimation of a pixel will use "3 * samples-count" Monte-Carlo realisations, one set of samples-count realisations for each X, Y and Z component of the CIE 1931 XYZ color space. By default, spp is set to 1.


Infinitely repeat the ground along the X and Y axis.


Infinitely repeat the clouds along the X and Y axis.

-m mie

Path toward a htmie(5) file defining the optical properties of water droplets.

-o output

File where htrdr writes its output data. If not defined, write results to standard output.

-T threshold

Optical thickness used as threshold criteria to partition the properties of the clouds. By default its value is 1.

-t threads-count

Hint on the number of threads to use. By default use as many threads as CPU cores.

-V x,y,z

Define the maximum definition of the acceleration data structure that partitions the cloud properties. By default the finest definition is the definition of the submitted htcp(5) file.


Make htrdr verbose.


Display version information and exit.


Render a clear sky scene, i.e. a scene without any cloud, whose sun is at zenith. The vertical atmospheric gaz mixture along the Z axis is described in the gas.txt file. The Mie data are provided through the file and the ground geometry is a quad repeated to the infinity. The camera is positioned at 400 meters high and looks toward the positive Y axis. The definition of the rendered image is 800 by 600 pixels and the radiance of each pixel component is estimated with 64 Monte-Carlo realisations. The resulting image is written to output excepted if the file already exists; in this case an error is notified, the program stops and the output file remains unchanged:

$ htrdr -D0,90 -a gas.txt -m -g quad.obj -R \
  -C pos=0,0,400:tgt=0,1,0:up=0,0,1 \
  -i def=800x600:spp=64 \
  -o output

Add clouds to the previous scene and use a more complex geometry to represent the ground. The ground geometry was carefully designed to be cyclic and can be thus repeated to the infinity without visual glitches. Use the -f option to write the rendered image to output even though the file already exists. Fianlly, use the htpp(1) command to convert the htrdr-image(5) saved in output in a regular PPM image [5]:

$ htrdr -D0,90 -a gas.txt -m -g mountains.obj -R -c clouds.htcp \
  -C pos=0,0,400:tgt=0,1,0:up=0,0,1 \
  -i def=800x600:spp=64 \
  -f -o output
$ htpp -o image.ppm output

Move the sun by setting its azimuthal and elevation angles to 120 and 40 degrees respectively. Use the -G option to enable the cache mechanism on clouds data. Increase the image definition to 1280 by 720 and set the number of samples per pixel component to 1024. Write results on standard output and convert the resulting image in PPM before visualising it through the feh(1) image viewer:

$ htrdr -D120,40 -a gas.txt -m -g mountains.obj -R -c clouds.htcp -G \
  -C pos=0,0,400:tgt=0,1,0:up=0,0,1 \
  -i def=1280x720:spp=1024 | htpp | feh -

Write into output the data structures used to partition the clouds properties. Use the csplit(1) Unix command to extract from output the list of the generated grids and save each of them in a specific VTK file whose name is cloud_grid_<NUM>*.vtk with NUM in [0, N-1] where N is the number of grids written into output:

$ htrdr -a gas.txt -m -c clouds.htcp -d -f -o output
$ csplit -f cloud_grid_ -b %02d.vtk -z --suppress-matched output /^---$/ {*}

Use mpirun(1) to launch htrdr on several hosts defined in the my_hosts file. Make the clouds infinite along the X and Y axis:

$ mpirun --hostfile my_hosts htrdr \
  -D120,40 -a gas.txt -m -g mountains.obj -R -c clouds.htcp -r \
  -C pos=0,0,400:tgt=0,1,0:up=0,0,1 \
  -i def=1280x720:spp=1024 \
  -f -o output


1. High-Tune: Gas Optical Properties file format -

2. OBJ file format -

3. VTK file format -

4. MPI specifications -

5. Portable PixMap -


Copyright © 2018-2019 CNRS, |Meso|Star> <>, Université Paul Sabatier <>. htrdr is free software released under the GPLv3+ license: GNU GPL version 3 or later You are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law.


csplit(1), feh(1), mpirun(1), htcp(5), htmie(5), htpp(1), htrdr-image(5)