Image.OCR.Input (image_ocr v0.2.0)

Normalises supported OCR inputs into a Vix.Vips.Image.t() and an associated raw pixel buffer suitable for handing to the Tesseract NIF.

Supported inputs:

%Vix.Vips.Image{} — used directly.
A binary that is an existing file path — loaded with Vix.Vips.Image.new_from_file/1.
A binary containing encoded image data (PNG, JPEG, TIFF, …) — loaded with Vix.Vips.Image.new_from_buffer/1.

Summary

t()

An accepted OCR input. See the moduledoc for resolution rules.

Returns a tightly-packed 8-bit pixel buffer for image, ready to feed to the Tesseract NIF.

to_vimage(image)

Loads input as a Vix.Vips.Image.t().

@type t() :: Vix.Vips.Image.t() | Path.t() | binary()

An accepted OCR input. See the moduledoc for resolution rules.

@spec to_pixel_buffer(Vix.Vips.Image.t()) :: {:ok, map()} | {:error, term()}

Returns a tightly-packed 8-bit pixel buffer for image, ready to feed to the Tesseract NIF.

The image is normalised in two ways:

Down-cast to 8 bits per band when needed.
Constrained to 1 band (grayscale) or 3 bands (RGB). 4-band RGBA images are flattened against an opaque white background; 2-band images are reduced to grayscale.

{:ok, %{pixels: binary, width: pos_integer, height: pos_integer, bytes_per_pixel: 1 | 3, bytes_per_line: pos_integer}} on success.
{:error, reason} on failure.

@spec to_vimage(t()) :: {:ok, Vix.Vips.Image.t()} | {:error, term()}

Loads input as a Vix.Vips.Image.t().