README

# rescribe.xyz/preproc package

This package contains various image processing methods which are
useful for preprocessing page images for OCR. It also contains
several commands in the cmd/ directory which can be used to
preprocess images directly.

This is a Go package, and can be installed in the standard go way,
by running `go get rescribe.xyz/preproc/...` and documentation
can be read with the `go doc` command or online at
<https://pkg.go.dev/rescribe.xyz/preproc>.

If you just want to install and use the commands, you can get the
package with `git clone https://git.rescribe.xyz/preproc`, and then
install them with `go install ./...` from within the `preproc`
directory.

## Commands

There are several commands in the cmd/ directory which are useful
in their own right as well as serving as examples of using the
package.

  - binarize     : binarises an image using the sauvola algorithm
  - pggraph      : creates a graph showing the proportion of black
                   pixels for slices through an image
  - preproc      : binarises and wipes an image
  - preprocmulti : binarises and wipes an image with multiple
                   binarisation ksize values
  - wipe         : wipes sections of an image that are outside an
                   area detected as content

## Bugs

The integral image operations don't produce exactly the same result
as their non-integral image counterparts. The difference is small
enough that it has little effect on the output images, but it ought
to be identical.

## Contributions

Any and all comments, bug reports, patches or pull requests would
be very welcomely received. Please email them to <nick@rescribe.xyz>.

## License

This package is licensed under the GPLv3. See the LICENSE file for
more details.

Documentation

Overview

preproc contains various image processing methods which are useful for preprocessing page images for OCR. It contains both library functions to incorporate into your own projects and standalone tools which can be used directly.

Index

• func BinToZeroInv(bin *image.Gray, orig *image.RGBA) (*image.RGBA, error)
• func IntegralSauvola(img image.Image, ksize float64, windowsize int) *image.Gray
• func PreCalcedSauvola(intImg integral.Image, intSqImg integral.SqImage, img image.Image, ...) *image.Gray
• func PreProcMulti(inPath string, ksizes []float64, binType string, binWsize int, wipe bool, ...) ([]string, error)
• func ProportionSlice(i SummableImage, x int, width int) float64
• func Sauvola(img image.Image, ksize float64, windowsize int) *image.Gray
• func VWipe(img *image.Gray, wsize int, thresh float64, min int) *image.Gray
• func Wipe(img *image.Gray, wsize int, thresh float64, min int) *image.Gray
• func WipeFile(inPath string, outPath string, hwsize int, hthresh float64, hmin int, ...) error
• type SummableImage

Functions

func BinToZeroInv

func BinToZeroInv(bin *image.Gray, orig *image.RGBA) (*image.RGBA, error)

BinToZeroInv converts a binary thresholded image to a zero inverse binary thresholded image

func IntegralSauvola

func IntegralSauvola(img image.Image, ksize float64, windowsize int) *image.Gray

Implements Sauvola's algorithm using Integral Images, see paper "Efficient Implementation of Local Adaptive Thresholding Techniques Using Integral Images" and https://stackoverflow.com/questions/13110733/computing-image-integral

func PreCalcedSauvola

func PreCalcedSauvola(intImg integral.Image, intSqImg integral.SqImage, img image.Image, ksize float64, windowsize int) *image.Gray

PreCalcedSauvola Implements Sauvola's algorithm using precalculated Integral Images

func PreProcMulti

func PreProcMulti(inPath string, ksizes []float64, binType string, binWsize int, wipe bool, wipeWsize int, wipeMinWidthPerc int, vWipeWsize int, wipeMinHeightPerc int) ([]string, error)

PreProcMulti binarizes and preprocesses an image with multiple binarisation levels. inPath: Path of input image. ksizes: Slice of k values to pass to Sauvola algorithm binType: Type of binarization threshold. binary or zeroinv are currently implemented. binWsize: Window size for sauvola binarization algorithm. Set automatically based on resolution if 0. wipe: Whether to wipe (clear sides) the image wipeWsize: Window size for wiping algorithm wipeMinWidthPerc: Minimum percentage of the image width for the content width calculation to be considered valid vWipeWsize: Window size for vertical wiping algorithm wipeMinHeightPerc: Minimum percentage of the image height for the content height calculation to be considered valid

func ProportionSlice

func ProportionSlice(i SummableImage, x int, width int) float64

ProportionSlice returns the proportion of black pixels in a vertical slice of an image starting at x, width pixels wide.

func Sauvola

func Sauvola(img image.Image, ksize float64, windowsize int) *image.Gray

Implements Sauvola's algorithm for text binarization, see paper "Adaptive document image binarization" (2000)

func VWipe

func VWipe(img *image.Gray, wsize int, thresh float64, min int) *image.Gray

VWipe fills the sections of image which fall outside the vertical content area with white, providing the content area is above min %

func Wipe

func Wipe(img *image.Gray, wsize int, thresh float64, min int) *image.Gray

Wipe fills the sections of image which fall outside the content area with white, providing the content area is above min %

func WipeFile

func WipeFile(inPath string, outPath string, hwsize int, hthresh float64, hmin int, vwsize int, vthresh float64, vmin int) error

WipeFile wipes an image file, filling the sections of the image which fall outside the content area with white, providing the content area is above min %. inPath: path of the input image. outPath: path to save the output image. hwsize: window size (width) for horizontal wipe algorithm. hthresh: threshold for horizontal wipe algorithm. hmin: minimum % of content area width to consider valid. vwsize: window size (height) for vertical wipe algorithm. vthresh: threshold for vertical wipe algorithm. vmin: minimum % of content area height to consider valid.

Types

type SummableImage

type SummableImage interface {
	image.Image
	Sum(r image.Rectangle) uint64
}