Filtering 3D viewer: phase 1

[pberkes]

We need to add a "Color by:" dropbox on the histogram that controls what is displayed in the histogram and the 3D viewer. By default, the dropbox is going to contain "CT" (i.e., the original data).

Values in the dropbox are given by a list of Filter objects that have a name (to be displayed in the dropbox) and a function, filter, that takes the 3D volume data and return a new 3D array with the same shape and potentially new data.

The output of the Filter object is used to compute a new histogram, and to update the content of the 3D view.

Things we need to care about:

• [ ] The dropbox responds to changes to the list of filters
• [ ] Filters have an updated Event trait; the histogram listens to that trait and when it is fired it calls the filter again to recompute its output
• [x] Update/add tests for VolumeData
• [ ] Refactor the masking functionality into a VolumeFilter subclass. See: https://github.com/enthought/ensemble/issues/46#issuecomment-147073792

[jwiggins]

Can this be thought of as a generalization of the volume masking feature?

[dmsurti]

@pberkes The implementation details have changed. Please refer to https://github.com/enthought/ensemble/pull/48.

We have a VolumeFilter class which has a filter method to transform the volume data's raw_data.

VolumeData has a VolumeFilter instance, which when set will apply the transform on the `raw_data1.

[pberkes]

@jwiggins it could, but we will want to filter (including transforming the values) and mask (e.g. with the cylindrical mask) at the same time

[dmsurti]

@jwiggins @pberkes Looking at _prepare_data (https://github.com/enthought/ensemble/blob/add-data-filters/ensemble/volren/volume_data.py#L143), this could be the possible refactors:

• ImageFilter(VolumeFilter) implements _image_data_from_array with spacing attribute.
• ResampleFilter(VolumeFilter) implements _resample_data
• MaskFilter(VolumeFilter) implements _apply_mask with mask_data attribute. MaskFilter will also have an ImageFilter instance as _apply_mask uses _image_data_from_array.

Also the above refactors seem to point to VolumeFilter as just a data transformation filter, more suitably named DataFilter if one were to implement as above.

OTOH, I think all this is an overkill? Thoughts?

[pberkes]

While I see the appeal of having a whole visualization pipeline, I do think that it would be overkill given what we are trying to accomplish: given resampled, masked data, we want the user to be able to use different filters to color the result.

[jwiggins]

What do you mean by "color the result"? The color is being generated by the transfer function.

[pberkes]

Bad wording: what I mean is that the content of the histogram and the displayed values are from the filtered data.