Abstract

       The use of Virtual Reality is very spread nowadays in many sides and domains. Although many VR headsets and hardware products are available, the creation of VR content can be challenging and time-consuming, given that, a lot of code should be written to generate or render objects and scenes in VR in a quiet good quality. Besides, high knowledge on graphics APIs and pipelines is necessary to ensure smooth rendering of the content and avoid potential issues. However, there is a solution can be a boost and speed up the process of VR content creation by using existing scenes or objects and scanning them and so the content creators can consider them as assets in VR.

       Moreover, Image Based Rendering can be a suitable technique for complex and large VR scenes because, it is based on capturing and processing images of the scene, rather than generating a 3D model from scratch. In this paper, we will investigate and analyze this new rendering technique based on the Unstructured Lumigraph Rendering (ULR). And this will be done using an Unity3d Game Engine based package called ColibriVR. Furthermore, we will test the performance of this technique using the very famous Stanford Bunny model and interpret the results, which will come out from our VR Headset. Additionally, we will provide in this paper benchmarks of Test cases using this specific 3D Model with different datasets, so that, we can analyze the Image based rendering using ULR as a blending method in this package and see, how far this package can be useful for VR content rendering.

Conclusion

In summary, we can say that we answered our two research questions ??, which we proposed at the beginning. Firstly, we explained and showed how we can create a data set from a 3D model using ColibriVR acquisition component. Secondly, we simplified our testing Stanford Bunny model into 3 other levels of details with fewer vertices and triangles. In the next step, our focus was shifted to render these models in VR using ColibriVR. Hence, we showed how to do that by using the Processing and Rendering component. In this part of our approach, we faced many challenges, which we fixed by adjusting the study package. In addition and after fixing the issues, we started analyzing and evaluating the results and
the performance of our PICO NEO 3 VR headset when rendering using the Unstructured Lumigraph Rendering (ULR) technique. After that, our main goal was to observe the rendering quality of the image we received from our headset. In that part, we noticed the large amount of ghosting artifacts and reduced them by adjusting some package parameters (Depth Correction).

Overall, our approach may help the content creators to get more information about using the image based rendering in ColibriVR with ULR.
In addition, we can confirm from this evaluation that it is possible to make optimizations or adjustments to this package to make it fit each specific goal. Thus, we can suggest this package to the non-experts, who don’t have any computer science knowledge and want to create content in VR.

Thesis:

Slides: