Parallelizing Over “Theta” - 2024.2 English - XD100

Vitis Tutorials: AI Engine Development (XD100)

Document ID
XD100
Release Date
2024-12-06
Version
2024.2 English

An alternative strategy partitions the Hough Transform such that each tile sees the full image but computes only a partial transform over a subset of the $\theta$ range. This also leads to a linear reduction in compute, but does not achieve any input bandwidth reduction. This scheme benefits from a linear reduction in the tile histogram storage. Collect the histogram outputs from each tile, but there is no extra compute workload involved with this approach.

figure