implement Alternativeimage-based processing #48
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Via `AnalyseLayout` and `OSD`, tesseract can determine the skew angle for images. The new wrapper applies this to pages and regions. It is not clear yet how to save the estimated skew angle in PAGE XML. Cf. PRImA-Research-Lab/PAGE-XML#9
- base all processors on AlternativeImage
- make all processors create a parameter MetadataItem
- make all processors create output file names
from the input files, and use .xml extension for PAGE
- introduce a `common` module along the lines
of the ocropy wrapper (but without
ocropy-specific segmentation), i.e. functions
to be moved into core:
- polygon_mask
- rotate_polygon
- image_from_page
- image_from_region
- image_from_line
- image_from_word
- image_from_glyph
- save_image_file
- bbox_from_points
- points_from_bbox
- xywh_from_bbox
- bbox_from_xywh
- points_from_polygon
- in crop:
- set textord_tabfind_find_tables=0 (because with
table detection, the hinge often gets confused
with a table column)
- if a Border already exists, warn that it will
be overwritten
- if TextRegions already exist, calculate their
common extent and warn it will be ignored
- use PSM.SPARSE_TEXT instead of PSM.AUGO (so
no images regions creep into neighbouring pages)
- ignore regions which are empty after binarization
- ignore regions with tiny width or height (< 30px)
- add a padding to the result on all sides (4px)
- do not annotate a (wrong) Border if no regions
have been found
- in deskew:
- convert skewing angle from radians to degrees,
and mind the direction (clockwise in PAGE, but
mathematically positive in Pillow) and map to
the numeric interval (-179,180)
- add orientation (+90/180/270) to skewing angle
- also rotate the raw image of the page/region
(expand and fill with white) and store as file;
reference in METS (under OCR-D-IMG-DESKEW) and
in PAGE (as AlternativeImage, with appropriate
comments)
- annotate writing direction and textline order
in PAGE too
- use OSD (DetectOrientationScript) in addition to
layout analysis (AnalyseLayout/Orientation), with
confidence thresholds (>= 10):
ensure that orientation is consistent between both
(and in case of conflict, use the former), also
annotate primary script; init appropriately
(i.e. load "osd", use legacy OEM and AUTO_OSD)
- on region level, process TableRegions as well
- change default operation_level to region (because
we still cannot annotate orientation on page level)
- in segment_region:
- add parameter `find_tables` (default: true) to allow
disabling table detection (textord_tabfind_find_tables=0),
so they can be analysed into independent text/sep regions
- add parameter `overwrite_regions` (default: true) to
allow enabling removal of any existing text regions
- unconditionally remove any existing non-text regions
and reading order groups
- cover PT.VERTICAL_TEXT (as TextRegionType) and PT.TABLE
(as TableRegionType)
- use BlockPolygon (if present) to annotate polygon outline
in Coords – but comment away, because patch against tesserocr
segfaults awaits merge
- add parameter `crop_polygons` (default: false) to enable:
retrieve the raw region image along the (internal) polygon
outline, store image as file, and reference in METS (under
OCR-D-IMG-CROP) and in PAGE (as AlternativeImage)
- in segment_line, add parameter `overwrite_lines` (default: true) to
allow enabling removal of any existing text lines
- in segment_word, add parameter `overwrite_words` (default: true) to
allow enabling removal of any existing words
- new processor binarize:
- operate on page, region or line level
- retrieve cropped, raw page/region/line image, then enter
PSM.AUTO/SINGLE_BLOCK/SINGLE_LINE, and run layout analysis
on the image, retrieve the binary image for RIL.BLOCK/TEXTLINE
store image as file, and reference in METS (under OCR-D-IMG-BIN)
and in PAGE (as AlternativeImage)
- improve docstrings
- remove redundant locale workaround from config
(already in __init__)
- version 0.2.2 → 0.2.3
This was referenced Jun 28, 2019
Merged
wrznr
suggested changes
Jul 3, 2019
wrznr
reviewed
Jul 3, 2019
|
|
||
| TOOL = 'ocrd-tesserocr-binarize' | ||
| LOG = getLogger('processor.TesserocrBinarize') | ||
| FILEGRP_IMG = 'OCR-D-IMG-BIN' |
There was a problem hiding this comment.
This can only be a temporary solution (cf. OCR-D/spec#117).
wrznr
approved these changes
Jul 3, 2019
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #26 (superseding #27) and #33 – the latter of course leaves some open points to be discussed in the spec. The idea and the functions in
common(for adoption in core) are the same as for ocropy.This also supersedes #34 (which I based on, but I was not allowed to force-push there after clean rebasing).