despeck is a Ruby gem that helps you remove unwated stamps or watermarks from scanned images/PDFs, primarily prior to OCR.

Its image processing operations are based on libvips via the ruby-vips Ruby-bindings.

It can be used to:

• detect uniform watermarks from a series of images,

• output a watermark pattern file (image, mask) that describes a watermark pattern, and

• remove a specified watermark pattern from input images regardless of the location of the watermark on these images.

Assumptions on input:

• The input may be a single image, or a PDF of multiple pages of images

• In the case of multiple pages, not all pages may have the watermark

• The input images are assumed to be purely monochrome text-based.

• The watermarks are colored and are located in a similar location within all the pages. For example, if the watermark is a GREEN SQUARE PATTERN, for all the pages that contain this mark, despeck will attempt to detect this pattern and remove them

Quick start

gem install despeck

Functionality

despect has three functions:

1. recognize a common watermark from a set of input images. All of the input images for detection should contain at least one of this mark.

2. output a file that specifies a particular watermark pattern (and the color of it)

3. remove a specified mark (from a "watermark pattern file" provided by step 2) from a series of input images. These input images may or may not contain this watermark.

Usage (library)

TODO: under development.

Usage (command-line)

Detection

The command shall do the following.

1. The command shall take the input of a PDF / image(s)

2. The command outputs a file that contains the detected watermark(s)

This is the sample help output for despeck extract.

$ despeck extract -h

Usage:
  despeck extract [options] <inputfile> <outputfile>

    [-s sensitivity]
      An integer value from 0 to 100 to indicate sensitivity of the watermark
      detection algorithm.

    [-w watermark-color-hint-in-RGB]
      One or more RGB 8-bit values in hexadecimal (0x00 to 0xFF), separated by
      commas for each RGB channel, to indicate the approximate color of the
      watermark.

      The significant bits (0 to 8) should be indicated after the full
      color hexadecimal value. Example: "FF,AA,88/6" to indicate that the last 2
      bits of this color value are to be ignored.

    <inputfile>
      The input file(s)

    <outputfile>
      The file to output the watermark image to.

    If multiple watermark candidates are detected, multiple watermark images

Removal

1. The command shall take the input of a PDF / image(s) and the watermark image

2. The command outputs a file that does not contain the watermark

Internally, it:

1. Investigates each page and detect which ones has a watermark.

2. For pages that do not contain the mark, just output it to the output file.

3. For the pages that contain the watermark, remove the mark and output to the output file.

4. Output the final PDF that all pages have the watermark removed.

This is the sample help output for despeck remove.

$ despeck remove -h

Usage:
  despeck remove [options] <inputfile> <watermarkfile> <outputfile>

    This command removes the watermark specified in <watermarkfile> from
    <inputfile>, and outputs the resulting image(s) in outputfile.

    [-s sensitivity]
      An integer value from 0 to 100 to indicate sensitivity of the watermark
      removal algorithm.

    <inputfile>
      The input file(s)

    <watermarkfile>
      The watermark file

    <outputfile>
      The file to output the resulting images that have the specified watermark
      removed.