Feature Extraction

Building the Cantata Workspace

1. Use the image pieces.viff for this experiment.
2. Label the image with the Labeling operator.
3. Display it with a random colortable using the Autocolor operator.
4. If the image looks noisy, filter it using a closing operation, i.e., a dilation followed by a erosion.
5. Use the Shape Analysis operator. The labeled image is the first input image parameter. Select the Invariant Moments. There are several outputs from these operators. The ASCII File output gives a readable printout of the feature value for each object. The VIFF Statistics File is a binary file with WIDTH and ELEMENTS. The width dimension is the same as the number of objects and the elements dimension is 61, where each element represents a particular feature measured by the operator. Look at the help page of the Shape Analysis operator to see the correspondence to the feature and the position in the element dimension.
6. Use the Extract operator to extract a particular feature. Element position 42 corresponds to first invariant moment.
7. Plot the extracted file to visualize the feature measured of each object.

Exercises

• The Shape Analysis operator numbers the objects starting from 0. The Labeling operator starts from 1. The background is labeled as 0. What can you do so that the plot appears consistent with the labeled image?
• Repeat the same experiment with another feature.