RGBA bytes owned independently of caller PDF input and engine memory.
OptionaldpiEffective resolution in pixels per inch, as the raster appears on the page, when known. Core sets it for page renders and image regions; providers may use it to tune recognition.
Eight-bit red/green/blue/alpha channel order.
Input raster height in pixels.
Optionallayoutsparse when the raster holds scattered words (handwriting, forms, labels) rather than
blocks of text, as core asks when a page yielded no text otherwise; recognizers may use a
sparse-text mode for it or ignore it.
Bytes per row; the bundled Tesseract provider requires width × 4.
Input raster width in pixels.
Owned RGBA8 working raster passed to an OCR provider. Provider geometry must refer to these input pixels.