Introduction
In our previous article about Photography & Simulation Comparison, we focused on presenting our photograph-rendering comparison process. We shed light on the meticulous process and protocols we employ to perform this comparison. This critical step in our workflow ensures the accurate representation of product appearance: faithful rendering of the aspect of your product necessitates a thorough understanding of optical properties and their proper modelling in our software. In the present article, we extend this methodology to validate material models on end-products, emphasizing the significance of precise 3D geometry and optical measurements for rendering complex geometries and materials.
Material modelling:
To perform a render of an object in our software OceanTM we introduce a light emitter, a 3D model of the product geometry and a material model. This model is based on the measured optical properties. These experimental measurements are usually performed on multi-thickness chips, as shown in Fig. 1, top. Total and diffuse reflectance, as well as transmittance, are measured on each section of the sample using an integrated sphere (Fig. 2). Performing the characterization on different thicknesses allows us to test the measurements reproducibility and to evaluating possible thickness-dependence. Based on these measurements, we extract optical parameters and build a material model in OceanTM.

Figure 1: Plastic chip - Right view

Figure 1: Plastic chip - Perspective view

Figure 1: Plastic chip - Front view

Figure 2: Integrated sphere - Total reflectance, total transmittance, diffuse transmittance
We performed integrated sphere measurements on the green and grey chips shown on the photographs of Fig. 3, top. Based on this experimental characterization we extracted optical coefficients and built a material model in OceanTM. We used our calibrated lightbooth environment to validate it through the photo-render comparison process. To do so, we performed a simulation in OceanTM: we placed the sample in our virtually reproduced lightbooth and positioned the instrument as our calibrated camera in our photography room. The simulated samples are shown on Fig. 3, bottom. Visual inspection of the photographs and renders allows us to validate the material model: we obtain an excellent matching of the samples’ appearance.
For more details on materials measurement for rendering projects, see our article “Introduction to Materials Measurement”.

Figure 3: Photograph of a grey chip

Figure 3: Photograph of a green chip

Figure 3: Simulation of the grey chip. Simulation generated with Ocean™.

Figure 3: Simulation of the green chip. Simulation generated with Ocean™.
Photo-render comparison of products:
We can use our lightbooth for several purposes, ranging from optical property estimation (stay tuned for a future article on this topic), model validation to product vizualisation in a controlled environment. Here, expanding beyond simple chips, we applied our photo-render comparison process to products with complex geometries. Using the validated material models on chips, we simulated a bottle’s appearance (Fig. 4) for which we wish to obtain a virtual visualization.
First, we simulated the appearance of this container made of the green material introduced in the previous section. Figure 4 shows the photo of the physical bottle on the left and the render of the bottle on the right.. As can be noticed here, while the green chip was faithfully reproduced, the color of the physical product and the rendered bottle are slightly different, despite an accurate material model.
This discrepancy highlights that the accuracy of the 3D CAD model is a crucial factor for simulating the end-product appearance. Indeed, in the 3D model used, the flatness of the top part of the bottle is not faithfully reproduced (red rectangle), leading to a different distribution of light in this region. While many complex optical phenomena contribute to this appearance discrepancy, a simple one is illustrated on Figure 4. According to Lambert’s cosine law, the irradiance on a surface is proportional to the cosine of the angle between the light direction and the surface normal.
Considering a simple uniform lighting from above of the bottle, we thus expect a lower irradiance on the top surface of the right container (Figure 4). Additionally, the physical sample displays thickness variations inherent to the production process. Such fluctuations are not accounted for in the geometrical model. The combination of these variations and the translucency of the green material leads to a different appearance.
This highlights that to obtain the most precise rendering, we need an accurate 3D model. This double check approach allows us to detect those types of inaccuracies and avoid disappointment before processing with the physical sample.



Figure 4 : Slight geometry differences between the object and the provided 3D CAD leading to appearance discrepancies. Simulation generated with Ocean™.
We also tested the second, grey-colored material model for rendering of the bottle. The physical product is shown in Figure 5, left. Simulating the appearance of the bottle reveals that the experimental characterization performed does not provide sufficient information to build a material model, adequately capturing the precise surface finish of the product. Therefore, we tested different optical models to simulate these subtle effects more accurately. Each model is based on both the optical measurements performed and the addition of an extra function to account for the glossy aspect of the end-product’s surface. For each model considered, we performed a photograph-render image comparison. In Figure 5, the left half of each image corresponds to the photograph and the right half is the rendered image. Based, on this analysis we determined that the material model employed to produce the top left image most faithfully captures the appearance of the provided sample.
This highlights that to obtain the most precise rendering, we need an accurate material model, that can be calibrated using our precise photograph-render comparison process.

Figure 5: Various simulations illustrating the differences in the appearance of the end product as a function of the optical models used. Simulations generated with Ocean™.
Product Visualization:
We then used our controlled lightbooth environment and calibrated material models for comparing the appearance of existing bottles with virtual counterparts. Virtual bottles refer to those that have not been physically produced but can be modelled using optical measurements and material models. The results of such simulations are shown on Fig. 6, where the first and third bottles are those introduced in the preceding sections, while the second and fourth are virtual representations.
Moreover to forsee the products in their natural usage environment, we can also conduct in-situ simulations. To do so, we design a scene based on customer specifications. For example, in the second picture in Figure 6, we rendered the bottles displayed on a bathroom shelf, lighted from the ceiling. This in-situ previsualization step is crucial to capture all the complexities of the product’s appearance. This in-situ simulation reveals subtle effects that were not visible under the homogeneous lighting of our controlled lightbooth environment.


Figure 6 : Simulations in controlled environment (left) and in-situ (right). Simulations generated with Ocean™.
Beyond plastics
In this study, the photo-rendering comparison was primarily used as a validation tool, but also as an evaluation protocole that helped defining the appropriate surface finish of the end-product, for which no physical characterization was available.
There are other use cases where experimental characterization of the material is impossible, such as in the modeling of faceted gems. The peculiar shape of faceted gems makes it impossible to characterise them accurately. Additionally, since each gem is unique, reusing an existing material that has already been measured won’t provide accurate results. In this case, photo-rendering comparison proves to be an essential tool for achieving accurate optical simulations.
The photo–render process allows to iterate on the optical model, starting from the literature data of the absorption of the gemstones, giving us more information about the intensity. Our process has proven to be reliable in various use cases, making it trustworthy to use it to provide data.


Figure 7: Comparison of photograph and simulation (generated with Ocean™) of gems.
As discussed in a separate article, geometry is not the only parameter to consider when predicting the appearance of a gemstone.
Conclusion
In summary, our exploration of photorealistic rendering and virtual prototyping highlights the critical role of accurate material models, 3D geometry rendering and controlled Lightbooth environments. From simulating complex geometries to comparing optical measurements and rendering products, our approach ensures the accuracy of results. The photo-render comparison not only allows us to validate materials models, but also helps to evaluate surface finishes, as shown in the bottle simulations. In-situ visualization permits capturing products’ complexity under different lighting conditions.
Beyond plastics, our study extends to challenging materials such as faceted gemstones, where experimental characterization is sometimes impossible. In such cases, photo-rendering comparison proves to be an indispensable tool for achieving accurate optical simulations. As we navigate this intersection of technology and aesthetics, our comprehensive methodology emphasises the need for calibrated material models for accurate rendering.
Stay tuned for more insights into our innovative virtual prototyping processes in future articles, as we explore different use cases and push the boundaries of accurate optical simulation.
All the simulations displayed in this article are generated with Ocean™.
Responses