Files
avdb-py/INDEXING.md
T
2026-10-04 18:24:18 -04:00

1.5 KiB

INDEXING

Overview

Each image is processed into multiple down-sampled images (powers of 2).

Each level is analyzed as

  • tile-sets (lower resolution stages)
  • gridded histograms (higher resolution stages) or
  • or not at all (highest resolution stages/too much detail)*

- orig
- detail/no analysis
- detail/no analysis
- gridded histogram
- gridded histogram
- gridded histogram
- tile-set
- tile-set
- tile-set

  • gridded-histogram X 81
    • 9 X 9 X 48 bin Lab histogram
  • gridded-histogram X 25
    • 5 X 5 X 48 bin Lab histogram
  • gridded-histogram X 9
    • 3 X 3 X 48 bin Lab histogram
  • tile-set X 9
    • ~300 X 9
  • tile-set X 5
    • 80 X 5
  • tile-set X 1
    • 20 X 1

COL LABELS

Mapping tiles to analysis data:

  • tile_id
  • img_id
  • level (resolution)
  • tilex
    • logical tile x
  • tiley
    • logical tile y
  • x
    • upper left @ histogram levels (in original resolution)
    • center @ tile-set levels
  • y
    • ...
  • x0
    • upper left @ histogram levels (in down-sampled resolution)
    • center @ tile-set levels
  • y0
    • ...
  • w
    • width of histogram region
    • or width/size of tile-set
  • h
    • height of histogram region
    • or width/size of tile-set (redundant with w - option 2)
  • foreign key to histograms table
  • foreign key to tile-sets table

Reverse mapping: analysis data to tiles:

  • foreign key to histograms table
  • tile id

OR

  • foreign key to tile-data table
  • tile-id

ORIGINAL_RESOLUTION <-> DOWN_SAMPLED_RESOLUTION

  • point
  • rectangle
  • area