Distant Hills and Trees

These elements transcend the horizon line in the matte painting (Fig 1) so applying this to a flat planar geometric surface would appear to be the logical approach. If there was some visible curvature of the hills then there would be some merit in considering applying the texture to the inside, or indeed outside surface of a primitive shape (such as a cylinder), deforming a Card or indeed using a more complex geometric shape

Fig 1: Shows the hills and trees beyond the meadow in 3D space

Once the decision is taken to use a Card as the target geometry for the element, I was able to compare three approaches to applying the texture.

I started by assigning the texture directly to the Card UV’s. In the 3D view this looks promising as the texture appears to present correctly (Fig 2). The first task involved establishing the required position for the Card in Z space, aligning the Card on its Y axis so it sits at the correct height relative to the ground and finally scaling the card so the element is the correct size.

Fig 2: Texture applied directly to the card UV’s

It is helpful to work with two viewers; one presenting the scene in 2D and the other in 3D (Fig 3). However despite this it is difficult and time-consuming to achieve an exact match with the original matte painting.

Fig 3: Reconfigured software interface to allow a 2D and 3D view of the scene

The second approach is, rather than working with the Transform properties on the Card, to use its optical attributes. The premise is that, if the card and camera share the same focal length and aperture, then we can use the Z attribute to move the Card in Z space and the optical attributes with scale and align the Card relative to the camera view. Fig 4a shows the properties panel for the shot camera with attributes set. Fig 4b shows how the focal length and the horizontal aperture values from the shot camera could be applied to the corresponding properties in the Card

Fig 4a: Camera properties with attributes set	Fig 4b: focal length and horizontal aperture attributes from the camera are applied to the card

This should therefore allow the Card to be repositioned in Z space but without the need to manually position and scale the Card using the transform attributes.

In this project, using this approach proved to be unsuccessful. This is because the Z attribute scales the Card from the scene origin (world space zero) so it will therefore only present the correct view to the shot camera if this is also placed at the scene origin. (The Foundry documentation). In this case the initial scene geometry (the horizontally orientated Card for the ground texture) had been constructed from the origin and the camera moved back on the Z axis to frame the artwork.

It would of course be possible to recreate the camera move so the hero frame of the animation actually resided at world space zero and then reposition the ground elements and sky relative to this. However this would be a significant amount of work, almost tantamount to starting the project afresh, and cannot be justified for such small gains.

The third approach would be to project the texture onto the Card rather than assigning it to the UV’s. The methodology here is identical to that described when projecting the ground texture. The ‘hero’ frame presents the true perspective of the element so presents accurately when seen from the shot camera. Providing the Card is set to the correct depth in Z, then the texture is projected into the correct position and at the correct scale.

Fig 5: A second camera, in a fixed position along the shot camera’s motion path is used to project the texture onto the geometry

Application of the two banks of trees required the same approaches and the same broad themes for analysis. From this point, use of the optical attributes on Cards is discounted due to the shot camera not being at the scene origin.

The comparison is therefore limited to assigning the textures directly to the UV’s of their respective target geometry or applying the texture using camera projection. Again both approaches were viable but the UV approach required more work manually transforming the geometry and the result was approximate rather than exact.

Placement of the Cards in depth could be achieved using visual cues from the matte painting. The left side trees align to the back part of the ground terrain and the right side trees align to the front of the distant hills.

Fig 6: Shows position and rotational transforms to the geometry to align more logically based on visual clues in the image