If the Dimension Organization is TILED_FULL, the position of frames in a multi-frame image needs to be computed. The current implementation of the compute_plane_position_slide_per_frame() function performs this computation in an inefficient manner by calling compute_plane_position_tiled_full() for each frame, which internally constructs a PixelToReferenceTransformer object for every invocation. This PR improves the performance of the computation by constructing a PixelToReferenceTransformer object once and then re-using the object repeatedly. I tested this on a larger VL Whole Slide Microscopy Image instance and was able to speed up the computation from 2.38 seconds to 1.65 seconds.
If the Dimension Organization is
TILED_FULL
, the position of frames in a multi-frame image needs to be computed. The current implementation of thecompute_plane_position_slide_per_frame()
function performs this computation in an inefficient manner by callingcompute_plane_position_tiled_full()
for each frame, which internally constructs aPixelToReferenceTransformer
object for every invocation. This PR improves the performance of the computation by constructing aPixelToReferenceTransformer
object once and then re-using the object repeatedly. I tested this on a larger VL Whole Slide Microscopy Image instance and was able to speed up the computation from 2.38 seconds to 1.65 seconds.