getLines: Find and Determine Vertical and Horizontal Line Segments in...

In duncantl/Rtesseract: Interface to the tesseract OCR system

Find and Determine Vertical and Horizontal Line Segments in an Image

Description

The functions getLines() and findLines() use leptonica Pix functions to identify vertical or horizontal lines in an image and then extract the coordinates of these segments, combining close segments as appropriate. findLines(), by default, processes the image to leave only the lines and getLines() uses this to get the line coordinates.

Usage

getLines(pix, hor = dims[2]* 0.02, vert = 5, lineThreshold = 0.1, fraction = 0.5,
         gap = 0.02, asDataFrame = FALSE, ..., asIs = is(pix, "AsIs"),
         horizontal = hor > vert, dims = dim(pix))
findLines(pix, hor = dims[2]* 0.02, vert = 5, asLines = TRUE, invert = !asLines,
          erode = c(3, 5), threshold = 210, convertTo8 = GetImageDims(pix)[3] > 8,
           dims = dim(pix))

Arguments

pix	the image, as a Pix-class object, or coercable to one, e.g., the name of an image file
hor, vert	the horizontal and vertical dimensions defining a window used to search for lines. Larger area windows mean a potential line has to occupy more pixels to be considered a line. This reduces the false positives. If asIs is TRUE, horiz and vert are not used.
lineThreshold	the proportion or number of pixels in the binary image matrix that must be black to be considered a line. Smaller values allow detecting short line segments. Too small means false positives.
fraction	when collapsing rows/columns in the matrix to a single "line" in the image, the proportion of the cells that have to be black for the corresponding cell in result to be black.
gap	the number of pixels below which two segments are considered one, i.e., when the end of one segment is within gap pixels of the start of the other segment.
asDataFrame	a logical value that controls whether the result is returned as a data.frame or a matrix.
...	additional arguments passed to findLines
asIs	indicates if the image passed via pix has already been processed to contain only the lines (TRUE). Otherwise, pix is passed to findLines and we extract the lines from the resulting image.
horizontal	a logical value indicating whether we are looking for horizontal or vertical lines. This can be useful if the pix value is the preprocessed image that contains only the lines. In this case, we don't have to specify horiz and vert but just horizontal.
asLines	a logical value that controls ....??
invert	a logical value that controls whether we invert the resulting binary image, turning black to white and vice versa, only used if asLines is TRUE.
erode	a numeric/integer vector with 2 elements, passed to pixErodeGray
threshold	an integer value used for thresholding the gray-scale pixels to create a binary, black and white image.
convertTo8	controls whether the Pix is automatically converted to 8 bit, if it is not already in that format.
dims	the dimensions of the image - rows and columns

Value

a list of matrices, each containing 4 columns giving x0, y0, x1, y1 coordinates or, if asDataFrame, we combine the list into a single data frame with the same columns.

Author(s)

Duncan Temple Lang

References

http://www.leptonica.com/line-removal.html

See Also

pixGetPixels

Examples

f = system.file("images", "SMITHBURN_1952_p3.png", package = "Rtesseract")

p1 = pixRead(f)
p1 = pixConvertTo8(p1)

bin = pixThresholdToBinary(p1, 150)

angle = pixFindSkew(bin)
p2 = pixRotateAMGray(p1, angle[1]*pi/180, 255)

h = findLines(p2, 101, 1, TRUE, erode = integer())
plot(h)

 # Compute the end points of the line segments
hlines = getLines(h, asIs = TRUE, horizontal = TRUE)
hl = do.call(rbind, hlines)
apply(hl, 1, function(x) lines(x[c(1, 3)], nrow(p2) - x[c(2,4)], col = "red"))

 # Here we allow a larger gap and also fewer of the cells in a group of
 # rows need to have a black pixel for the corresponding pixel in the
 # line to be black.
hlines = getLines(h, asIs = TRUE, horizontal = TRUE, gap = 250, fraction = .2)
hl = do.call(rbind, hlines)
apply(hl, 1, function(x) lines(x[c(1, 3)], nrow(p2) - x[c(2,4)], col = "red"))


vlines = getLines(p2, 1, 101)

duncantl/Rtesseract documentation built on Sept. 8, 2024, 8:38 a.m.