1.0 Taking images for modelling

Please note that images must be either Windows Bitmaps (.bmp) or JPEG files (.jpg).  The majority of digital cameras use the JPEG standard for storing images.

If you wish to use Viewpoint Estimation to have Viewing Directions calculated automatically then place the object on a piece of paper lying on a flat surface. The paper acts as a Viewpoint Target and must be one of the following sizes,

Note : It does not matter exactly where the object is placed on the paper, although it may help in understanding this document if you place your object with it's front facing one of the long sides of the rectangle, as these are refered to elsewhere as the Front and Back edges of the Viewpoint Target.

Now take images of the object using any fixed-focal-length (no zoom) digital camera, OR a zoomable-digital-camera with the zoom fixed at a constant level.  In other words if your camera has zoom, leave it set at the same level in every image you take.  If possible make a note of the zoom settings used.

If you are using a Viewpoint Target DO NOT MOVE THE OBJECT BETWEEN ONE IMAGE AND THE NEXT, MOVE THE CAMERA. It is important that the relative position of the model to the Viewpoint Target remains a constant within the project.

The quality of the finished model depends on the suitability of the images for modelling.  Which images to take depends largely upon the shape of the object being modelled, and taking a good set of images takes some experience.  The following general guidelines should be adhered to,

1.1 Advice for taking images

  1. The object should be as large as possible in each image (by taking the image close to the object), subject to 2 below.
  2. The whole object and paper should be visible in each image
  3. Attempt to capture as much of the surface of the object as possible, by taking images from a diverse set of viewing positions.
  4. In almost all cases it is advisable to take 3 or more images, and if only three are used then these should be taken from i) above, ii) from one side and iii) from the front of the object.  Please note that images may be taken from anywhere, the viewpoints described above need not be followed exactly.
  5. In addition to these recommended viewpoints, images should be taken of the object so that most aspects of it's shape have been seen.  For example, if building a model of a cup, the hole in the handle of the cup should be visible in at least one of the images. 
  6. Between 5 and 10 images are usually enough to build a model which approximates the object.  Objects are never modelled exactly, but including more images, in general, improves the object model. 

A good strategy when building a model is as follows,

1.2 A simple strategy for deciding on the best views to use

  1. Take a set of three views of the object, approximately from top, side and front.
  2. Build a model with the model appearance set to No Texture to get a shape-only model.
  3. View the model using the viewer (Using the Tools->View 3D Model menu option), decide from which directions this model looks incorrect and take extra images from approximately these viewing positions.
  4. Repeat this process from point 2 above, until a good shape approximation is made.
  5. Build a textured model by setting the model appearance to Texture Map, and view this.  If it has any blank patches on it's surface, try to take an image which views this area and include this in the project.

This process may be seen as an iterative procedure of shape refinement.  An other way of refining the shape of models is described later, in which the viewer program  is used to refine the shape of the object - without the need to take extra views with a camera.

1.3 Adding Images to a Project

Start Real2Virtual modeller and select the menu option, Project->Views->Add/Remove Images...

The following dialog box is shown,

type the name of an image file in the Add File edit box - or alternatively select the Browse... button to select one or more files from a standard interactive file-browser. 

Click on the Add Above button to add the images to the project, above the current list position, or Add Below to add below the current position. If this is the first image pressing either has the same result.

The Remove button may be used to delete Image files or Blank-Views from the project, either now or at any later time. When the Remove button is pressed the currently selected list item is removed from the list, and is put in the Add File edit box. This allows it to be inserted in another position, or simply discarded. In this way the order of the images in the project may be altered at any time.

When the images have all been added select Close, alternatively to discard changes select Cancel

Se here about image properties 

1.4 Browsing Project Views

You can browse through the project views using the image browser buttons on the Image Toolbar.  These are shown below,

Use the arrow buttons to move to the First, Previous, Next or Last view.  Also you may select a project view by name, from the drop-down-box on the image toolbar.

To magnify a view select the Zoom In button, and then select a point somewhere in the view.  This point acts as a "focus of attention".  Zoom in further by repeatedly selecting points.

Alternatively the magnification may be changed using the drop-down on the image toolbar.

To restore normal magnification press the No Zoom button.

1.5 Creating Blank Views

A blank view is one which may only contain silhouettes.  It does not have an associated image.  To define a blank view select Project->Files->Add Blank View. and fill in the following dialog,

To fit in with other views in the project whose size is measured in Pixel units, a blank view also has a size defined in pixels.  A unique label is also required to refer to the view.

When the details are complete select OK and a blank view is added at the end of the image list.  Select the view and press the right mouse button to see it's properties.

Real2Virtual Modeller Documentation Version 0.3
© Real2Virtual 2000-2002.  All rights reserved.