By myst6re.

Camera

The format is almost the same as Final Fantasy VII. Differences:

  • The zoom can be present twice
  • There may be several cameras, in this case the camera structure is repeated several times, and sizeof(each struct) = 0x28 (with zoom twice for each)

Structure

Offset Size Data
0x00 3 × int16 Camera X axis (4096 = 1.0)
0x06 3 × int16 Camera Y axis
0x0C 3 × int16 Camera Z axis
0x12 int16 Copy of the Z axis’ z
0x14 3 × int32 Translation
0x20 4 bytes Unused
0x24 uint16 Zoom
0x26 uint16 Zoom again (only in the 0x28-byte multi-camera layout)

Projecting a field position onto the background

A field position P (walkmesh / SET3 coordinates) goes to camera space with the axes as the rows of a rotation matrix, then to the screen with the zoom as the projection distance:

C = (axes / 4096) · P + translation
screen_x = C.x × zoom / C.z
screen_y = C.y × zoom / C.z

The screen position is relative to the background’s origin, the point where the tiles X and Y are 0. Projecting the whole walkmesh of bghall_1 this way draws it exactly over the floor, stairs and bridge of the background.