unpaper - User Documentation

This document describes the concepts and features of unpaper.

Table of Contents

Basic Concepts
Sheets and Pages
Input and Output Image Files
Layouts and Templates
Processing Multiple Files
Masks
Borders
Size Values
Image Processing Features
Blackfilter
Noisefilter
Blurfilter
Grayfilter
Deskewing (Auto-Straightening)
Processing Order

Basic Concepts

The terminology to describe unpaper makes heavy use of the paper metaphor, because the software is mainly intended for post-processing scanned images from printed paper documents.

Sheets and Pages

The very basic object unpaper operates on is a sheet. A sheet is an initially blank image in the computer's memory. Think of a sheet as an initially empty piece of paper on which something will be printed later.

To do something useful with a sheet, you will at least want to place one page onto a sheet. A page is a logical unit of a document which takes up a rectangular area on a sheet. In the most simple case, one sheet carries exactly one page,  in other cases (e.g. when using a double-page layout) there can be multiple pages placed on one sheet.


Figure 1: Sheets and Pages

Input and Output Image Files

unpaper can process either double-page layout scans or individually scanned pages. It is up to the user's choice whether an image-file carries a single page or a whole sheet with two pages. The program can be configured to either join individual image-files as multiple pages onto one sheet, or split sheets containing multiple pages into several output image-files when saving the output.

By default, unpaper places one input image-file onto a sheet, and saves one output image-file per sheet. Alternatively, the number of input or output image-files per sheet can be set to two using the --input-pages 2 or --output-pages 2 options.

If two image-files are specified as input, they will successively be placed on the left-hand half and the right-hand half of the sheet.


Figure 2: Input Files

In the same way, if two image-files are specified as output, the sheet will be split into two halfs which get saved as individual files.


Figure 3: Output Files

The default value both for --input-pages and --output-pages is 1.

File Formats

The image-file formats accepted by unpaper are those of the PNM-family: PBM, PGM and PPM. This ensures interoperability with the SANE tools under Linux. The simplicity of the PNM file-formats also allows unpaper to be developed as a pure standalone program without any dependencies on external packages or libraries, because the PNM file-formats are very easy to handle.

In order to operate on images with a different file-format, several image file-format converters are available which can be used to convert to PNM before processing, and afterwards convert back from PNM to the desired file format. Look e.g. at pngtopnm / pnmtopng, tifftopnm / pnmtotiff, jpegtopnm / pnmtojpeg, or consider using the all-purpose image file-type converter convert from the imagemagick package.

Layouts and Templates

Built-In Layout-Templates

Layouts are the linking concept between physical sheets and logical pages. A layout determines a set of rectangular areas at which pages (or other parts of content) appear on a sheet. The most common and simple layouts generally used are the single-page layout (one page covers the whole sheet), and the double-page layout (two pages are placed on the left-hand-side and the right-hand-side of the sheet).

unpaper provides basic layout templates for the above types, both in horizontal and vertical orientation of the sheet. Thus, there are 4 layout templates built in:
  1. single
  2. single-rotated
  3. double
  4. double-rotated

Figure 6: Available Layout Templates

A layout template is chosen by using the option --layout, e.g.

unpaper --layout double input%03d.pbm output%03d.pbm.

Choosing a template with the --layout option is equivalent to specifying a set of other options, e.g. setting --mask-scan-point or --pre-rotate / --post-rotate.  In order to combine a template with other options, make sure that the more specific options appear behind the --layout option, in order to overwrite the template settings.

Note: A layout is completely independent from the number of image-files used as input or output. That means, you can either specify --layout double together with a single input image-file (in cases where the input image-file already contains two scanned pages in a double-page layout), or use it together with an --input-pages 2 setting, in order to join two individually scanned pages on one sheet.

Complex Layouts

Besides the built-in fix templates, any kind of complex layout can be handled by manually specifying either mask-scan-points using the --mask-scan-point option, or setting masks at fix coordinates using the --mask option. Both the --mask-scan-point and the --mask option may occur any number of times, in order to declare as many masks in the layout as desired. See below for a further explanation on masks.

Processing Multiple Files

In many cases, especially when post-processing scanned books, there will be several input image-files to process in sequence within a single run of unpaper, and several output image-files to be generated. Processing of multiple files in a batch job is supported through the use of wildcards in filenames, e.g.:

unpaper (...options...) input%03d.pbm output%03.pbm

This will successively read images from files input001.pbm, input002.pbm, input003.pbm etc., and write output to the files output001.pbm, output002.pbm, output003.pbm etc., until no more input image-files with the current index number are available.

Using a wildcard of the form "%0nd" will replace each occurrence of the wildcard with an increasingly counting index number, by default starting with 1 and counting up by 1 each time another files gets loaded. n denotes the  number of digits that the replaced number string is supposed to have, and the 0 requests leading zeros. Thus "%03d" will get replaced with strings in the sequence 001, 002, 003 etc. This way, a sequence of images named e.g. input001.pgm, input002.pgm, input003.pgm... can be specified. There are two seperate index counters for input and output files which get increased independently from each other.

Wildcards in filenames are also useful when combining a sequence of individual pages onto double-page layouted sheets, or when splitting double-page layouted sheets into individual output files. When using two input or output image-files (by specifying --input-pages 2 or --output-pages 2) the index number replaced for the wildcard will generally not be the same as the sheet number in the processing sequence, but will grow twice as fast.

The following example will combine single-page image-files onto a double-page layout sheet:

unpaper -n --input-pages 2 singlepage%03d.pgm output%03d.pgm

This joins the input images singlepage001.pgm and singlepage002.pgm on output001.pgm, singlepage003.pgm and singlepage004.pgm on output002.pgm, and so on. Note that due to the use of option -n (short for --no-processing), the images are simply copied onto the left-hand half and the right-hand half of the sheet without any processing regarding layout, mask-detection etc.

Using multiple input image-files by setting --input-pages 2 is independent from any layout possibly specified with the --layout option. However, in order to use unpaper's post-processing features for more than simply joining two image-files to one, you will most likely want to combine the use of --input-pages 2 with the --layout double option, as in:

unpaper --layout double --input-pages 2 (...other options...) singlepage%03d.pgm output%03d.pgm


Figure 4: Sequence of multiple input images

On the contrary, it is also possible to split up a sheet into several image-files when saving. The following line would be used to split up a sequence of double-page layouted sheets into a sequence of single-page output images, including full image processing (applying masking, deskewing, border-aligning etc., see below), in order to make sure that the pages in the double-page layout are really placed fully on the left-hand half and the right-hand half of the sheet before the sheet gets split up:

unpaper --layout double (...options...) --output-pages 2 doublepage%03d.pgm singlepage%03d.pgm


Figure 5: Sequence of multiple output images

By default, processing of multiple sheets starts with sheet number 1, and also with input and output image-files number 1. unpaper will run as long as input image-files with the current index number can be found. If no more input files are available, processing stops.

Adjusting Indices

In order to start with a different sheet index, the --start-sheet option can be set. Accordingly, setting --end-sheet specifies a fix sheet number that will the last one to be processed, even if more input-files are available.

Using --sheet, a single sheet or a set of specific sheet numbers to be processed can be specified. For example:

unpaper --sheet 7,12-15,31 --input-pages 2 (...options...) input%03d.pgm output%03d.pgm

This would generate the output-files output007.pgm, output012.pgm, output013.pgm, output014.pgm, output015.pgm and output031.pgm, reading input from the same files as if a whole sequence of sheets and pages starting with index 1 had been processed, i.e. reading the files input013.pgm and input014.pgm for sheet 7, input023.pgm and input024.pgm for sheet 12, and so on.

To prevent some sheets from being processed (i.e., remove them from the sequence), the option --exclude can be used. Note that this is different from option --no-processing or -n, which still would generate the output files but without applying any image processing to them.

The input and output index numbers to start with can be adjusted using the options --start-input and --start-output. These values apply to the wildcard replacement in filenames only and are independent from the sheet numbering. In other words, setting these options specifies an offset at which the file numbering starts relative to sheet 1. For example:

unpaper --input-pages 2 (...options...) --start-input 7 input%03d.pgm output%03d.pgm

These settings would cause the input-files input007.pgm and input008.pgm to be used for sheet 1, input009.pgm and input010.pgm for sheet 2, and so on. The default value for both options is 1.

File-Sequence Patterns

More sophisticated file-sequence patterns can be specified using the --input-file-sequence or --output-file-sequence options. In cases where the input files are named after a pattern like e.g. left01.pbm, right01.pbm, left02.pbm, right02.pbm etc., the use of --input-pages 2 together with --input-file-sequence left%02d.pbm right%02d.pbm will load to the desired images. The index counter with which the wildcards in the filenames get replaced is increased every time the file-sequence pattern is iterated through, it will not be increased after each single replacement of a wildcard.

Note that it would also be possible to use file-sequence patterns of different lengths than the number of pages per sheet. In case an input file-sequence like e.g. a%d.pbm b%d.pbm c%d.pbm is specified together with --input-pages 2, the input image-files used for the first sheet would be a1.pbm and b1.pbm, the input image-files used for the second sheet would be c1.pbm and a2.pbm (!), for the third sheet they would be b2.pbm and c2.pbm, and so on. It's up to the user whether it makes sense to use file-sequence patterns of different length than the corresponding number of input image-files or output image-files per sheet.

Specifying a filename as the very last argument on the command-line is equivalent to using --output-file-sequence <file> (a sequence of length 1), specifying a filename as the last-but-one argument on the command line is equivalent to using --input-file-sequence <file>.

Inserting Blank Content

Input file-sequences may be forced to use completely blank images at some index positions. The --insert-blank option allows to specify one or more input indices at which no file is read, but instead a blank image is inserted into the sequence of input images. The input image that would have been loaded at this index position in the sequence will be used at the following non-blank index posisiton instead, thus the following indices get shifted to make room for the blank image inserted.

The --replace-blank option also allows to insert blank images into the sequence, but it suppresses the images that would have been loaded at the specified index positions and ignores them. No index positions get shifted to make room for the blank image.

Masks

Masks are rectangular areas on a sheet that are affected by several of the processing steps unpaper performs. Although there may be as many masks on a sheet as desired, in most cases it will be useful to operate with either one or two masks per sheet only. A single-page layout would operate on only one mask covering the whole page, a double-page layout would make use of two masks, one placed somewhere in the left-hand half of a sheet, the other somewhere in the right-hand half.

Automatic Mask-Detection

Masks can be set directly by specifying pixel coordinates using the --mask option, but in most cases it is desirable to detect masks automatically. Automatic mask-detection allows input images to contain content which is not perfectly placed at fix areas, but probably differs slightly in position from sheet to sheet (which is usually the case when books are scanned or photocopied manually).

Automatic mask-detection uses a starting point somewhere on the sheet called mask-scan-point, which marks a position estimated to be somewhere inside the mask to be detected. (When detecting masks that cover a whole page, it is useful to place the mask-scan-point right in the center of the sheet's half on which the page appears.) Beginning from the mask-scan-point, the image content is virtually scanned in either the two horizontal directions (left and right), or the two vertical directions (up and down), or all four directions, until no more dark pixels are found which means an edge of the mask is considered to have been found.


Figure 7: Mask-Detection

Several parameters control the process of mask-detection. At first, mask-scan-points to start detection at get specified either using the --layout option (which automatically sets one mask-scan-point for single-page layouts, and two mask-scan-points for double-page layouts) or manually with the option --mask-scan-point.

Mask-detection is performed by the use of a 'virtual bar' which covers an area of the sheet under which the number of dark pixels is counted.
The 'virtual bar' is moved towards the directions specified by --mask-scan-direction. (Those directions not given via --mask-scan-direction will use up the whole sheet's size in these directions for the detected result.)

While moving the 'virtual bar' the number of dark pixels below it is continually compared to the amount that has been counted at the very first position of the 'virtual bar' above the mask-scan-point when detection started. Once the amount of dark pixels drops below the relative value given by --mask-scan-threshold, mask-detection stops and an edge of the mask is considered to have been found.


Figure 8: 'Virtual Bar' for Mask-Detection

The width of the 'virtual bar' can be configured using the --mask-scan-size option, the length of it by setting --mask-scan-depth. Adjusting the 'virtual bar's' width can help to fine-tune the process of mask detection according to the content that is being scanned. The wider the 'virtual bar' is, the more tolerant the detection process becomes with respect to small gaps in the content (which is e.g. needed if a page is made up of multiple columns). However, if the 'virtual bar' is too wide, detection might not stop properly when a mask's edge should have been found.

Figure 9: Mask-Scan-Threshold

Mask-detection can be disabled using the --no-mask-scan option, optionally followed by the sheet numbers to disable the filter for.

Mask-Centering

Masks that have been automatically detected or manually set will be used for several further processing steps. At first they provide the basis for properly centering the content on the corresponding page area on the sheet.

This allows to automatically correct inprecise positions of page content in scanned sheets and shift the content to a normalized position. Especially when processing multiple pages, this leads to more regular positions of pages in the sequence of resulting sheets.

Figure 10: Mask-Centering

Mask-centering can be suppressed using --no-mask-center, optionally followed by the sheet numbers to disable the filter for.

Borders

Unlike masks, borders are detected by starting at the outer edges of the sheet (or left/right halfs of the sheet, in a double-page layout), and then scanning towards the middle until some content-pixels are reached. Again, a 'virtual bar' is used for detection, the width of which can be set using the option --border-scan-size, and the step-distance with which to move it by setting the option --border-scan-step. The option --border-scan-threshold determines the maximum absolute number of pixels which are tolerated to be found below the 'virtual bar' until border-detection stops and one edge of the border area is considered to have been found.


Figure 11: Border-Detection

Border-Aligning

Borders serve two different purposes: First, the area outside the detected border on the sheet will be wiped out, which is another mechanism to clean the outer sheet boundary from unwanted pixels.

Second, a detected border can optionally be aligned towards one edge of the sheet. Border-aligning means shifting the area inside the border heading towards one edge of the sheet. The direction towards which to shift the border gets specified with the option --border-align. Additionally, a fix distance to the edge is kept, which can be set via --border-margin.
This way, it can be assured that e.g. all pages of a scanned book regularly start 2 cm below the upper sheet edge.


Figure 12: Border-Aligning

Note that border-aligning is not performed by default, it needs to be explicitly activated by setting the option --border-align to one of the edge names top, bottom, left or right, and by setting --border-margin to the desired distance which is to be kept to this edge.

Use --no-border-scan to disable border-detection, or --no-border-align to prevent border-aligning on specific sheets, both optionally followed by the sheet numbers to disable the filters for.

Size Values

Whenever an option expects a size value, there are three possible ways to specify that:
Using one of the last two ways, length measurements get internally converted to absolute pixel values based on the resolution set via the option --dpi. If the default of 300 DPI should be changed, this option must appear on the command line before using a length measurement value. --dpi may also appear multiple times, e.g. if the size values of the output image(s) should be based on a different resolution than those of the input file(s).

Note that using the --dpi option will have no effect on the resolution of the image-files that get written as output. (The PNM format is not capable of storing information about the image resolution.) The value set via --dpi will only have effect on unpaper's internal conversion of length measurements to absolute pixel values when size values are specified using length measurements or size names.

Image Processing Features

Blackfilter

Sometimes it is desirable to automatically remove large black areas which originate from bad photocopies or other optical influences. The blackfilter can help to find large areas of black and wipe them out automatically.


Figure 13: Blackfilter

Be careful with pictures in scanned documents, especially with diagrams. Some diagrams intentionally contain large areas of dark color, which might be affected by automatic wipe-out of the blackfilter. In order to prevent actual page content from being wiped out, the option --blackfilter-scan-exclude allows to specify areas on the sheet which should not be taken into account by the blackfilter. When using one of the default layout templates set via the --layout option, the inner area of each page will automatically be excluded from black-filtering.


Figure 14: Blackfilter Details

The blackfilter can be disabled by the option --no-blackfilter, optionally followed by the sheet numbers to disable the filter for.

Noisefilter

The noisefilter removes small clusters of pixels ("noise") from the sheet.
The maximum pixel-size of clusters to be removed can be set via --noisefilter-intensity. This value must not be chosen too high in order not to remove relevant elements of page content, e.g. normal text-points ("."). As a consequence, this option might have to be adjusted on images with a low scan resolution.


Figure 15: Noisefilter

Disable with --no-noisefilter, optionally followed by the sheet numbers to disable the filter for.

Blurfilter

The blurfilter removes "lonely" clusters of pixels, i.e. clusters which have only very few other dark pixels in their neighborhood.


Figure 16: Blurfilter

The size of the neighborhood to be searched and the amount of other dark pixels accepted in the neighborhood below which the area gets wiped out can be adjusted with the options --blurfilter-size, --blurfilter-step and --blurfilter-intensity. Additionally, --blurfilter-step also determines the step-size with which the neighborhood-area is moved across the image while filtering.


Figure 17: Blurfilter Details

Disable with --no-blurfilter, optionally followed by the sheet numbers to disable the filter for.

Grayfilter

The grayfilter removes areas which are gray-only, that means it wipes out all those areas which do not contain a maximum relative amount of non-dark pixels. The size of the local area the grayfilter operates on can be set using --grayfilter-size, and the granularity of detection is controlled via --grayfiter-step. The maximum relative amount of non-dark pixels that are still considered to be deletable can be set using --grayfilter-threshold.

Be careful with the grayfilter when processing color scans, because any bright color might be considered as gray and be wiped out. It might be a good idea to disable the grayfilter when processing color scans.

Disable with --no-grayfilter, optionally followed by the sheet numbers to disable the filter for.

Deskewing (Auto-Straightening)

The deskewing performed by unpaper is actually a rotation to automatically straighten rectangular content areas on the sheet. It is applied to any mask that has been found during mask-detection or that has been set directly via the --mask option.


Figure 18: Deskewing

The algorithm that detects which rotation angle is to be used works the better, the more regular and solid the edges of the area's content are.
It works as follows:
A 'virtual line' is moved from the outside of one edge inside the rectangular area. This happens several times, gradually changing the rotation of the 'virtual line'. (Called 'virtual', because there is of course no visible line drawn in the image.) The algorithm will count the number of dark pixels along the line as it is virtually moved.

Some parameters control the the size of the 'virtual line' and its movement:
--deskew-scan-size: the height/width of the 'virtual line' used for scanning (the length of the line at rotation angle 0)
--deskew-scan-range: the absolute value of degrees between the negative and positive value of which the line will be rotated (i.e., the default value 5.0 will cause the 'virtual line' to be rotated in several small steps between -5.0 degrees and 5.0 degrees).
--deskew-scan-step: the step size with which to iterate between the bounds set by --deskew-scan-range (I.e., a value of 0.1 will lead to the virtual line being successively rotated with 0.0, 0.1, -0.1, 0.2, -0.2, 0.3, -0.3 ... degrees.)


Figure 19: Rotation Detection

At each of these rotation steps, the following is done: The rotated 'virtual line' gets moved (again 'moved virtually') towards the center of the rectangular area on which detection gets performed. Movement is performed pixel by pixel, it starts with the line completely outside the rectangular area, not yet reaching inside the area. At each movement-step, the number of dark pixels covered by the virtual line is counted and is accumulated as the total sum of dark pixels. For each rotation angle at which this is done, the maximum difference in the accumulated sum of dark pixels occurring between a previous movement-step and the next one gets calculated. The rotation angle for which this maximum difference becomes maximal will be the rotation angle detected for deskewing.

The relative amount of dark pixels to accumulate before shifting the 'virtual line' is stopped (and continued with the next rotation-step) is given by --deskew-scan-depth. This value is relative to the number of pixels that the 'virtual line' covers in total, i.e. for the default deskew-scan-size of 1500 and the default deskew-scan-depth of 0.66, shifting at each rotation step stops after 1000 dark pixels have been counted in sum (or, if not enough pixels are met, when the 'virtual line' has reached the center of the rectangular area).

Sometimes, trying out different deskew-scan-depth values, either lower than the default of 0.66, or higher, can noticeably increase detection quality. Which value is best is merely coincidental, depending on the shape of the outer edges of each very first character in each line of a text area.


Figure 20: Rotation Detection Details

The above described the detection process starting at one single edge of a 4-edged rectangular area (e.g. the left edge, as displayed in the above image). However, the overall rotation angle detection uses results of up to all four edges. Which edges to use can be specified by --deskew-scan-direction. The final rotation angle will then be the average value of all rotation angles detected at each edge. Usually, the individually detected values can be expected to be almost the same at each edge, if the rectangular area to be deskewed has a regular shape.  If, however, the individual values differ too much, it can be concluded that something went wrong with the detection, and no deskewing should be performed. (E.g., if the rotation at the left edge appears to be -0.5 degrees, but at the right edge results in 1.9 degrees, one should better not use the average value, because with that big difference something seems to have gone wrong with the detection.) So, before using the average of all individually detected values, their statistical standard-deviation is calculated, which is squareroot((a-average)^2 + (b-average)^2 + ...). If the standard-deviation among the detected angles exceeds the value specified by --deskew-scan-deviation, the total result is considered to be wrong and no deskewing is performed.

Deskewing can be disabled with --no-deskew, optionally followed by the sheet numbers to disable it for.

Processing Order

Processing of the filters and auto-corrections is performed in a fixed order according to the following sequence:
  1. load image file(s)
  2. perform pre-rotate, pre-mirror etc. actions on the individual input files (if specified)
  3. place on the sheet (multiple input-files are placed as tiles), auto-determine sheet size by the size of the input image-file(s) if not specified explicitly
  4. apply noisefilter and blurfilter to remove small bits of unwanted pixels
  5. apply blackfilter and grayfilter to remove larger areas of unwanted pixels
  6. detect masks starting from specified mask-scan-points
  7. perform deskewing on each detected or directly specified mask
  8. re-detect masks again to get precise masks after deskewing
  9. center masks on the corresponding page's area on the sheet
  10. perform border-detection
  11. align the detected borders
  12. save output image file(s), possibly perform post-rotate, post-mirror etc. actions on the individual output files before saving

Figure 21: Processing Order

Disabling processing steps

Each processing step can be disabled individually by a corresponding --no-xxx option (where xxx stands for the feature to disable, e.g. --no-grayfilter, --no-mask-scan etc.).
If such an option is followed by a sheet number, or a comma-seperated list of multiple sheet numbers, the filter gets disabled only for those sheets specified. Otherwise (if no sheet number follows), the filter is disabled for all sheets. Instead of specifying individual sheet numbers, also a range of numbers can be given, e.g. "10-20" to represent all sheet numbers between 10 and 20. Example:

unpaper (...options...) --no-blackfilter 3,15,21-28,40 (...)

This will disable the blackfilter on the sheets 3, 15, 21, 22, 23, etc. until 28, and 40.

Written by Jens Gulden 2006
Modifications under the GPL are welcome.