We target to run our application on low and mid-end mobile
phones. On such devices with low processing capabilities,
the percentage of RAM used during online query processing
becomes an important factor. The vocabulary and
the inverted index that we need for retrieval on the mobile
phone, consume most of the RAM during the query processing
stage. We reduce RAM usage by reducing the vocabulary
size without aecting the retrieval performance. This,
we call as vocabulary pruning.
Our approach in pruning the vocabulary is by removing
the visual words that are less important. We do this in