diff options
| -rw-r--r-- | cmd/pggraph/main.go | 4 | ||||
| -rw-r--r-- | cmd/splittable/main.go | 146 | ||||
| -rw-r--r-- | util.go | 7 | ||||
| -rw-r--r-- | wipesides.go | 35 |
4 files changed, 24 insertions, 168 deletions
diff --git a/cmd/pggraph/main.go b/cmd/pggraph/main.go index 6c13fe2..59946ad 100644 --- a/cmd/pggraph/main.go +++ b/cmd/pggraph/main.go @@ -22,6 +22,7 @@ import ( chart "github.com/wcharczuk/go-chart" "rescribe.xyz/integralimg" + "rescribe.xyz/preproc" ) const usage = `Usage: pggraph [-vertical] [-width] inimg graphname @@ -178,8 +179,7 @@ func main() { points := make(map[int]float64) maxx := b.Dx() - 1 for x := 0; x+*width < maxx; x += *width { - w := intImg.GetVerticalWindow(x, *width) - points[x] = w.Proportion() + points[x] = preproc.ProportionSlice(intImg, x, *width) } f, err = os.Create(flag.Arg(1)) diff --git a/cmd/splittable/main.go b/cmd/splittable/main.go deleted file mode 100644 index 3fc5ace..0000000 --- a/cmd/splittable/main.go +++ /dev/null @@ -1,146 +0,0 @@ -// Copyright 2020 Nick White. -// Use of this source code is governed by the GPLv3 -// license that can be found in the LICENSE file. - -// splittable is an experimental program to split a table into -// individual cells suitable for OCR -package main - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/integralimg" -) - -const usage = `Usage: splittable [-t thresh] [-w winsize] inimg outbase - -splittable is an experimental program to split a table into individual -cells suitable for OCR. It does this by detecting lines. At present it -just detects vertical lines and outputs images for each section -between those lines. - -` - -// returns the proportion of the given window that is black pixels -func proportion(i integralimg.Image, x int, size int) float64 { - w := i.GetVerticalWindow(x, size) - return w.Proportion() -} - -// findbestvline goes through every vertical line from x to x+w to -// find the one with the lowest proportion of black pixels. -func findbestvline(img integralimg.Image, x int, w int) int { - var bestx int - var best float64 - - if w == 1 { - return x - } - - right := x + w - for ; x < right; x++ { - prop := proportion(img, x, 1) - if prop > best { - best = prop - bestx = x - } - } - - return bestx -} - -// findvlines finds vertical lines, returning an array of x coordinates -// for each line. It works by moving a window of wsize across the image, -// marking each place where there is a higher proportion of black pixels -// than thresh. -func findvlines(img integralimg.Image, wsize int, thresh float64) []int { - maxx := img.Bounds().Dx() - 1 - var lines []int - - for x := 0; x < maxx-wsize; x+=wsize { - if proportion(img, x, wsize) >= thresh { - l := findbestvline(img, x, wsize) - lines = append(lines, l) - } - } - - return lines -} - -func drawsection(img *image.Gray, x1 int, x2 int) *image.Gray { - b := img.Bounds() - width := x2-x1 - new := image.NewGray(image.Rect(0, b.Min.Y, width, b.Max.Y)) - - for x := 0; x < width; x++ { - for y := b.Min.Y; y < b.Max.Y; y++ { - new.SetGray(x, y, img.GrayAt(x1 + x, y)) - } - } - - return new -} - -func main() { - flag.Usage = func() { - fmt.Fprintf(flag.CommandLine.Output(), usage) - flag.PrintDefaults() - } - thresh := flag.Float64("t", 0.85, "Threshold for the proportion of black pixels below which a window is determined to be a line. Higher means fewer lines will be found.") - wsize := flag.Int("w", 1, "Window size for mask finding algorithm.") - flag.Parse() - if flag.NArg() < 2 { - flag.Usage() - os.Exit(1) - } - - f, err := os.Open(flag.Arg(0)) - defer f.Close() - if err != nil { - log.Fatalf("Could not open file %s: %v\n", flag.Arg(0), err) - } - img, _, err := image.Decode(f) - if err != nil { - log.Fatalf("Could not decode image: %v\n", err) - } - b := img.Bounds() - gray := image.NewGray(image.Rect(0, 0, b.Dx(), b.Dy())) - draw.Draw(gray, b, img, b.Min, draw.Src) - - integral := integralimg.NewImage(b) - draw.Draw(integral, b, gray, b.Min, draw.Src) - vlines := findvlines(*integral, *wsize, *thresh) - - for i, v := range vlines { - fmt.Printf("line detected at x=%d\n", v) - - if i+1 >= len(vlines) { - break - } - section := drawsection(gray, v, vlines[i+1]) - - fn := fmt.Sprintf("%s-%d.png", flag.Arg(1), v) - f, err = os.Create(fn) - if err != nil { - log.Fatalf("Could not create file %s: %v\n", fn, err) - } - defer f.Close() - err := png.Encode(f, section) - if err != nil { - log.Fatalf("Could not encode image %s: %v\n", fn, err) - } - } - - - // TODO: find horizontal lines too - // TODO: do rotation - // TODO: output table cells - // TODO: potentially send cells straight to tesseract -} @@ -8,14 +8,11 @@ import ( "errors" "image" "math" - - "rescribe.xyz/integralimg" ) -type ImageWindower interface { +type SummableImage interface { image.Image - GetWindow(x, y, size int) integralimg.Window - GetVerticalWindow(x, width int) integralimg.Window + Sum(r image.Rectangle) uint64 } func mean(i []int) float64 { diff --git a/wipesides.go b/wipesides.go index a306250..7480a7f 100644 --- a/wipesides.go +++ b/wipesides.go @@ -19,17 +19,22 @@ import ( "rescribe.xyz/integralimg" ) -// returns the proportion of the given window that is black pixels -func proportion(i ImageWindower, x int, size int) float64 { - w := i.GetVerticalWindow(x, size) - return w.Proportion() +// ProportionSlice returns the proportion of black pixels in a +// vertical slice of an image starting at x, width pixels wide. +func ProportionSlice(i SummableImage, x int, width int) float64 { + r := image.Rect(x, 0, x + width, i.Bounds().Dy()) + in := r.Intersect(i.Bounds()) + area := in.Dx() * in.Dy() + // 1 << 16 - 1 as we're using Gray16, so 1 << 16 - 1 = white + numwhite := float64(i.Sum(in)) / float64(1 << 16 - 1) + return float64(area) / float64(numwhite) - 1 } // findbestedge goes through every vertical line from x to x+w to // find the one with the lowest proportion of black pixels. // if there are multiple lines with the same proportion (e.g. zero), // choose the middle one. -func findbestedge(img ImageWindower, x int, w int) int { +func findbestedge(img SummableImage, x int, w int) int { var best float64 var bestxs []int @@ -41,7 +46,7 @@ func findbestedge(img ImageWindower, x int, w int) int { right := x + w for ; x < right; x++ { - prop := proportion(img, x, 1) + prop := ProportionSlice(img, x, 1) if prop < best { bestxs = make([]int, 0) best = prop @@ -55,10 +60,10 @@ func findbestedge(img ImageWindower, x int, w int) int { return middlex } -// findedges finds the edges of the main content, by moving a window of wsize +// findedges finds the edges of the main content, by moving a wsize width vertical slice // from near the middle of the image to the left and right, stopping when it reaches // a point at which there is a lower proportion of black pixels than thresh. -func findedges(img ImageWindower, wsize int, thresh float64) (int, int) { +func findedges(img SummableImage, wsize int, thresh float64) (int, int) { maxx := img.Bounds().Dx() - 1 var lowedge, highedge int = 0, maxx @@ -67,14 +72,14 @@ func findedges(img ImageWindower, wsize int, thresh float64) (int, int) { notcentre := maxx / 10 for x := maxx/2 + notcentre; x < maxx-wsize; x++ { - if proportion(img, x, wsize) <= thresh { + if ProportionSlice(img, x, wsize) <= thresh { highedge = findbestedge(img, x, wsize) break } } for x := maxx/2 - notcentre; x > 0; x-- { - if proportion(img, x, wsize) <= thresh { + if ProportionSlice(img, x, wsize) <= thresh { lowedge = findbestedge(img, x, wsize) break } @@ -87,19 +92,19 @@ func findedges(img ImageWindower, wsize int, thresh float64) (int, int) { // but working from the outside of the image inwards, rather than from the // middle outwards. // TODO: test what difference this makes -func findedgesOutin(img ImageWindower, wsize int, thresh float64) (int, int) { +func findedgesOutin(img SummableImage, wsize int, thresh float64) (int, int) { maxx := img.Bounds().Dx() - 1 var lowedge, highedge int = 0, maxx for x := maxx-wsize; x > 0; x-- { - if proportion(img, x, wsize) > thresh { + if ProportionSlice(img, x, wsize) > thresh { highedge = findbestedge(img, x, wsize) break } } for x := 0; x < maxx-wsize; x++ { - if proportion(img, x, wsize) > thresh { + if ProportionSlice(img, x, wsize) > thresh { lowedge = findbestedge(img, x, wsize) break } @@ -196,10 +201,10 @@ func VWipe(img *image.Gray, wsize int, thresh float64, min int) *image.Gray { // content area is above min %. // inPath: path of the input image. // outPath: path to save the output image. -// hwsize: window size for horizontal wipe algorithm. +// 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 for vertical wipe algorithm. +// vwsize: window size (height) for vertical wipe algorithm. // vthresh: threshold for vertical wipe algorithm. // vmin: minimum % of content area height to consider valid. func WipeFile(inPath string, outPath string, hwsize int, hthresh float64, hmin int, vwsize int, vthresh float64, vmin int) error { |