Pre-Processing Examples

What follows are some examples of pre-processing and their results. Our aim is not to compare results, but to show some possibilities when basic operators are combined. Filters were implemented using vconv3d which performed a 3D convolution with 5x5x5 average mask. Segmentation was done by threshold, using vthresh3d. Interpolation was computed by vinter.


Interpolation - Filtering - Segmentation - Filtering

In this case was used linear between slice interpolation with gray method. On each step of this pre-processing are showed three renderings, using normal vectors estimated in voxel space (vvsnorm), image space (visnorm) and object space (vvsnorm), respectively.

After interpolation:

_ _

After first filtering:

_ _

After last filtering:

_ _


The following images are the same showed above, with their normal vectors filtered by 2D convolution using a 3x3x3 average mask.

After interpolation:

_ _

After first filtering:

_ _

After last filtering:

_ _


Segmentation - Interpolation - Filtering

In this case was used linear between slice interpolation with euclidean method and binary output. After interpolation are showed two renderings, using normal vectors estimated in image space and object space , respectively. Voxel space estimation can not be done because the scene is binary. After filtering, three renderings are showed as before, that is, using normal vectors estimated in voxel space, image space and object space.

After interpolation:

_______________________ _

After filtering:

_ _


The following images are the same showed above, with their normal vectors filtered by 2D convolution using a 3x3x3 average mask.

After interpolation:

_______________________ _

After filtering:

_ _


Segmentation - Filtering - Interpolation

In this case was used linear between slice interpolation with gray method. After interpolation are showed three renderings using normal vectors estimated in voxel space, image space and object space, respectively.

After interpolation:

_ _


The following images are the same showed above, with their normal vectors filtered by 2D convolution using a 3x3x3 average mask.

After interpolation:

_ _