This has been modified from the MSc Dissertation “Photo-realistic Reality: Focusing on artistic space at Çatalhöyük” (Cox 2010)
This post documents the research that was used to create the visualisations found on our gallery page. Before checking out this post you might want to familiarize yourself with a few other precursory posts, including the history behind the Shrine of the Hunters. If you wish to know more about research at Çatalhöyük, you can also find site reports from the current excavations here.
You can find part 1 of this series here.
Creating the wall paintings of F.V.I
The 3D Mural of the F.V.I paintings (Hi-rez version)
The main challenge in recreating the look of the Hunting Shrine of Level V was to recreate a series of wall paintings that were found in situ in the original excavations in the 1960’s by James Mellaart. After searching the site reports for any Munsell recordings for the wall paint, it was clear that the majority of recorded shades had been made in relation to soil colour, or clay based finds and plaster. The only application of it on visible paint (at the point of the creation of the model) were related to instances found on pottery. Inference into red values for paint were noted here, “Paint colours were generally red, distributed around 10R 4/6. ‘Red wash’ colours were more variable, possibly because the thickness of the application of the paint varies. There were two main groups, one which ranged from 10R 4/6 to 5YR 7/6 and another smaller group around 2.5Y 8/2-4.” (Last 1994) and another example of recorded colour for paint on pottery.
On reflection, rather than use these values, it was decided that finding a way to base all of the paint colours from the same context and not from varying areas of the site, would provide continuity to the work. Consequently, in Mellaart’s report (1965), the reconstructions compiled by Miss Raymonde Enderle Ludovic are specified as being the closest match by eye to the hues uncovered in situ as possible. Crucially, this meant that the recording was undertaken at a point when the paintings had been less influenced by degradation due to factors such as air and sunlight and it also means that any values came from the same direct source. In reality there is not much difference between this approach and a Munsell chart, as it was simply matching colour by eye on both occasions to a predefined selection. This was therefore chosen as the preferred route to attain a variety of colour information for the different painting shades.
Blend Mapping to split colour
Inside of 3DS Max there are multiple ways to create materials and in this instance, to provide future-proofing for the work and flexibility, it was decided that providing individual shaders for each colour would allow control over their physical properties (reflection, roughness etc), whilst also providing a framework with the ability to drastically alter, or test different eventualities. Inside of Mental Ray, the best way to achieve this result was to use a series of nestled blend maps which allows multiple materials to be split by a black and white input to differentiate where the separation occurs (Usually referred to as a mask), where pure white represents material 1 at its strongest intensity and pure black material 2. Greyscale amounts ranging between absolute white & black create the effect of a blend between the two choices. For the murals a series of masks were created in Photoshop from the paintings by Miss Raymonde Enderle Ludovic, of which an example can be seen below from painting one.
In situ shot of painting one (Mellaart 1965: Plate LII)
Recorded colour painting from the original excavation (Haydaroglu 2007)
A breakdown of the different paint layers
To create the panels, the masks were split into their respective paint colours and arranged to allow for their separation as noted below:
First panel: The mask used to split between the plaster and the painted areas. The black areas represent applied paint and the white the plaster.
Second Panel: This mask represents the areas that held pink colour and with the first panel, the two options in the blend map at this level were split between this panel and the next panel to break up the areas above into their respective areas. Anything that was not masked out by the black areas in the mask at this stage was coloured with the red ochre material.
Third Panel: This represents the black body hair areas of the men in the image painted in red ochre.
Fourth Panel: These areas are the black paint within the already masked pink areas in panel two. Not only are there defined black areas, but there are two separate textures, one for the black paint on red ochre seen above and one
An example of the blend. Material 1 & 2 held an Arch & Design shader, whilst the mask for each level was controlled by one of the black and white images above.
The ‘wrapper’ for each paint layer. This allowed separate versions of the paint to be created, with differing properties if required, providing a great deal of customisation.
A breakdown of the masks in action with blended materials. It is possible to see how each blend map is nestled within the next to create a chain of separation
Approaching the materials in this way opened up a level of control regarding the physical properties of the painted sections that could not be acquired using a simple bitmap image. It also meant that each colour could be solely covered by a single material that if needed could be re-routed back into various blend maps more than once. The ability to be self-sufficient in this way provided an incredibly beneficial platform on which to base future tests into visual properties detailed below, as each element can be individually edited when needed and hypothetically, it is therefore also possible to also test a wide range of possibilities for the paint layers, by just plugging new materials into the blend maps. A few examples of the possibilities are listed below:
Differing lighting conditions
Wetness
Roughness/Smoothness
Colour/Hue
Reflection/Refraction
Index of refraction falloff
Creating Texture information
Much of the detail for the underlying roughness and physicality of the paint layers came directly from high-resolution photographic and data capture from the site, specifically PTM maps that recorded features of paint application. An example of the hand print below shows a reference photograph and its normal details that were used to inspire a series of bump maps for the paint. Inside of 3DS Max, there are a number of ways to create the illusion of physical bumps/recesses on objects. One of those, bump mapping requires the use of a black and white map, where white represents high points and black, low. However, one of the draw backs of this technique is that it often requires direct light to be effective and due to the very limited amount of low light in the F.V.I model another more sophisticated method was required, normal mapping. This records a direction (left/right, up/down etc) into each RGB channel of the image and therefore these maps can simulate the profile of objects even in low light because they hold information that does not require direct light. This made normal mapping a perfect method for the highly indirectly light nature of the model and to create them the NDo plugin was used for Photoshop.
Hand print reference photograph/normals
Tiled normal map example created from reference
Breakdown of the basic paint shader using the PTM as a base
Plaster Detail & Props
Breakdown of the room plaster shader
The room UVWs unwrapped to prepare for painting
Breakdown of the floor shader
A selection of props with assorted maps
Dust shader
For the general plaster in the room, noise and composite maps were often used to implement random variety. To achieve a believable wear, the whole structure was unwrapped and taken into Photoshop, where custom dirt and degradation was painted using a colour palette taken from the existing photography and smoke and rain marks were manually created using the reconstructed house at the site as a reference. Often when creating very dusty, or uneven surfaces the FG/Highlights option was used in the Arch & Design shader, this sped up the process of the reflections in the room and also removed a few harsh artifacts. For almost every objects the scene reflective falloff was managed through the Arch & Design IOR system and varied based on material type, with the majority of surfaces falling between 1.5-2.0. Ambient occlusion was turned off to speed up the initial renderings and applied in post production using the techniques discussed here. Finally, to create the dust shader a standard material was used with an additive transparency and a face map and gradient ramps were used to create a spherical element that was then varied based on particle age to add variety to the lifespan of the dust. The resulting participating media sequence was then rendered out as a separate element and then layered onto the final shot.
References
Haydaroglu (ed.), 2007. Çatalhöyük: From Earth to Eternity. Yapi ve Kredi Ban- kasi. Istanbul.
Last, J. 1994, Pottery Report – Part II: Report on surface investigations at Çatalhöyük 1993-1994, in ÇATALHÖYÜK 2009 ARCHIVE REPORT, www.catalhoyuk.com (Accessed 02/10/2010)
Mellaart, J, Excavations at Çatal Hüyük, 1965, Forth Preliminary Report, In Anatolian Studies, Vol. 16 (1966), pp. 165-191, British Institute at Ankara: Ankara
