Acquiring full sphere images:


Two opposing fisheye images with 180° or larger fields of view are required. One approach is the Spherecam, a camera system that uses a pair of 220° Nikon fisheye lenses on a pair of Nikon F camera bodies.

An alternate approach is to take a pair of opposing photographs with 180° fisheye lenses. Suitable lenses include the Nikon 8 mm f2.8 lens and some of the Sigma 8 mm f4 lenses. Not all of the Sigma 8 mm lenses have the necessary 180° coverage.
 
 

Processing full sphere images


Once fisheye photo pairs have been photographed, they need to be combined in a computer. For the examples here, the combination is done with Adobe Photoshop, version 4.0. The first step is to form a cylindrical perspective intermediate image. From this image, the following images can be formed:

  1. VR compatible view in all directions -- similar to Apple QTVR and Omniview.
  2. Cylindrical format panoramic image -- similar to the image produced by a scanning panoramic camera.
  3. Spherical format hyperfisheye image -- similar to the image produced by a 360° fisheye lens.
The following section describes how to form the cylindrical intermediate image and then form a VR compatible view (option 1) that can be viewed with the RealVR viewer. The processing needed for options 2 and 3 are described in another article. Much of the following processing can be automated as an Action in Adobe Photoshop. Click here to download an action file for processing Spherecam photographs. This action file would need to be slightly modified for processing 180 degree fisheye images.

The following list are the Spherecam processing steps needed to form the cylindrical intermediate image:

  1. Scan film frames -- For the Nikon LS-1000 film scanner set image size 23 x 23 mm at 1350 or 2700 dpi. This minimizes unneeded additional interppolation steps. Center the scan area on to the circular fisheye image. Scan both images.
  2. Rotate both images to the same orientation -- Use (Image>Rotate Canvas>) options in Photoshop as needed to bring both images into the same orientation.
  3. Start Photoshop action sequence -- The following sequence of steps can be automated as a macro in Photoshop version 4. The action file assumes that the imagess were scanned at a 23 x 23 mm size at 1350 dpi.
  4. Convert image 1 to rectangular form -- Use the polar to rectangular conversion filter (Filter>Distort>Polar Coordinates) to convert the 1st imagee to a cylindrical projection. Use the Polar to Rectangular option in this filter.
  5. Increase the canvas size of image 1 -- Increase the image height from 0.908" to 1.481" when using a 220° fisheye lens (ratio of 360 to 220).
  6. Select image 2 and convert to rectangular form -- Use the polar to rectangular conversion filter to convert the 2nd image to a cylindrical projection.
  7. Rotate image 2 by 180 degrees -- This will align the lower and upper image segments.
  8. Copy image 2 to the clipboard -- The entire image should be selected.
  9. Close image 2 -- This image is no longer needed.
  10. Paste image 2 as a layer onto image 1 -- A new layer is created.
  11. Align layer 2 relative to layer 1 -- Image 2 is aligned relative to image 1 by using the offset filter (Filter>Other>Offset). Typical offsets are around 4450 pixels vertical and 0 pixels horizontal. The offset filter must be set to the wrap around mode.
  12. Blend transition between layer 1 and 2 -- Create a layer mask and blend between the images.
  13. Apply color corrections to both layers -- Apply any needed color corrections, both to make the transition region disappear and to correct the scene color balance andd gamma, etc.
  14. Merge the layers into a single image -- When the layer processing is complete, merge the 2 layers into a single layer.


At this point you have the cylindrical perspective intermediate image. Use the following additional processing to form a dynamically viewable VR image.

  1. Save the intermediate file as a JPG -- Convert the image to the desired pixel density and size and save as a jpeg file.
  2. Create a VRML control file -- The VRML control file can be created in Wordpad or another text editor. The JPG image file name needs to be imbedded in thee VRML file. The VRML control file should end with the file extension name ".ivr". The following is an example of the text in the ivr file assuming that the jpeg image name is "example.jpg".
  3. #VRML V2.0 utf8

    Vista {

    type "SPHERE"

    filename "example.jpg"

    }
    Click Here for an example.
     
     
  4. Install the RealVR viewer -- The RealVR viewer extracts dynamic views from the JPG image by using the ivr file to define the view properties. There are several types of RealVR viewers including stand alone, Netscape plug in and Internet Explorer plug in. These viewers can be downloaded at no cost from http://www.livepicture.com . You can test the viewer by double clicking on the ivr file that you created and associating it with the internet browser or stand alone RealVR viewer.


Home > Spherecam


© Copyright 1996, by Dan Slater, All Rights Reserved