The first row of images (Figure 15) displays little visible indirect lighting and focuses mostly on comparing the indirect diffuse bleed from the curtains onto the arches. Here it can be observed that the indirect diffuse light from VCT produces more blurred indirect color bleeding compared to the sharperlighting produced by Cycles. The light also bleeds back behind the curtains with VCT, while it sharply stops with Cycles render. This is one of the key problems with the VCT implementation. The individual voxels stored in the texture may end up thicker than thin geometry, thus creating a conflict between the occluded and directly illuminated sides.
These two sides will then end up averaged into the same voxel texture slot bleeding some light over to the occluded side (see Figure 16 for a more detailed example). Note that the flowers are also a bit darker in the Cycles render which is primarily due to unresolved issues with texture transparency displaying as black.
The second row of images (Figure 15) focuses more on displaying indirect lighting, emitted from the light silhouettes cast through the open windows by a distant light source. Here the differences become more apparent. Firstly, the floor and top left arches are too bright which is likely caused by the previously mentioned light leaking. In this case though, the leak begins only in higher mip-levels of the texture around the window edges. Another problem can be observed in the color that is bleeding from the curtains up against the pillars, which does not quite match. In the Cycles render the indirect light climbs a bit further up the pillar. This likely has to do with the cone configuration. In this case the cone offsets coupled with a 30° blind spot perpendicular to the normal could cause the curtain to be partially missed. Utilizing more cones with a smaller aperture tends to improve detail but comes with a sometimes-significant performance cost.