3dcgi.com - Home

Cool 3-D Graphics Technologies
3-D Cameras
3-D Displays
3-D Modeling
Image or Object Based Modeling
Conclusion and Links
Learn about 3-D graphics
3-d Graphics Links
3-d News
Computer Performance Tips

Add to My Yahoo!

3dcgi.com Monthly/Bi-Monthly Newsletter


Web 3dcgi

3-D modeling methods

Image or Object Based Modeling

Reductions in the cost of image-based modeling software (also known as photogrammetry), 3D cameras, and digitizers will make it easier than ever to create accurate models of existing objects. MicroScribe-3D

3D cameras and digitizers are good for small objects, but they are sometimes difficult to use for large objects, like buildings. High costs for the hardware have traditionally limited these technologies to big studios, but prices have been coming down.

The picture to the right is of a MicroScribe-3D digitizer that can be used to model objects up to 66" in size. This type of technology has been used in many applications, including Disney and Pixar's Toy Story. Watch the Ultimate Toy Box DVD set to see how Pixar used this technology.

Image-based modeling software has been around for a number of years. Recently the technology has matured enough to be considered for mainstream applications. With these programs no artistic talent is needed to create very detailed models. Just snap pictures of the object you wish to model, assign common reference points in software, and let the program analyze the photos to create a model. Image-based modeling works with any object.

[Update 2010] As time goes on software gets more advanced and it's now possible to create 3D models without assigning reference points. Agisoft PhotoScan and its free little brother StereoScan are two such programs. PhotoScan takes in multiple images as input, aligns them, and outputs a 3D model in one of many formats.

While image-based modeling technology can work with any object not all of the products do equally well. PhotoModeler, for example, works well with architectural objects, but it is not as efficient for modeling organic objects, like people.

This is because PhotoModeler requires users to manually assign reference points for areas that can be seen in multiple images. Each of these reference points form a vertex in the final model. Because organic objects require a lot more vertices than inorganic objects the sheer number of reference points needed to create a quality model increases development time dramatically. Also, it is often difficult to accurately identify points that are present in multiple images when there are no sharp edges to use for reference.

A couple of products use different methods to tackle the problem of modeling organic objects. Eyetronics ShapeSnatcher uses a grid, which is shined on the object from a slide projector, to automatically align the images and create a model. The grid is not part of the texture.
D Vision Works Limited
uses reference points to align the photographs, but the actual model is processed automatically. The user needs only to draw an outline around the object in the D Sculptor software to help make the processing easier.

The downside to this approach is D Sculptor seems to require more photographs, than other programs, to create a model. This isn't really much of a downside though because reference points for complicated organic objects would probably still take longer. Some types of objects will not work well with this method. You quickly can test how good this method is for your object by turning the model 360° and studying the outline of the model. If a lot of details are hidden by the outline then this method of image based modeling might not be right for you. Strata Foto3D and iModeller 3D are other applications that seem to use the masking method.

It should be mentioned that PhotoModeler and some other reference point modelers are so accurate for architectural models that they can actually be used to obtain detailed measurements of the objects. This makes these programs extremely useful for architects and engineers.

Some products are designed specifically to create models of faces and characters. Two are Cyberextruder and 3DMeNow. Cyberextruder creates a model from just one photograph while 3DMeNow requires two. 3DMeNow should give better quality models, but those from Cyberextruder should be just fine for games like Quake 3 and The Sims. 3DMeNow can also animate the model using one of 35 actions. The Pro version can output the model and animation data to programs like 3ds max and Maya. It even includes morph-targets for facial animation and lip-synching applications like LIPSinc VentriliquistT.

The products mentioned above have all been software packages that must be installed on your computer. Scanetica has taken a different approach by creating a web service. It doesn't require any special equipment or software and for now at least (Feb. 2009) it's free to use. The service can handle a wide range of subject sizes from satellite imagery to tiny fossils.

Ironically, image or object based modeling gives new life to traditional art forms, like sculpting. Once an object is sculpted it can easily be converted into a computer model.

Click here for a list of vendors and links, or read on to the end of the modeling article.



3dcgi.com Home Page