diff options
Diffstat (limited to 'cmd')
-rw-r--r-- | cmd/binarize/main.go | 82 | ||||
-rw-r--r-- | cmd/preproc/main.go | 94 | ||||
-rw-r--r-- | cmd/preprocmulti/main.go | 105 | ||||
-rw-r--r-- | cmd/splittable/main.go | 143 | ||||
-rw-r--r-- | cmd/wipe/main.go | 59 |
5 files changed, 0 insertions, 483 deletions
diff --git a/cmd/binarize/main.go b/cmd/binarize/main.go deleted file mode 100644 index a844982..0000000 --- a/cmd/binarize/main.go +++ /dev/null @@ -1,82 +0,0 @@ -// Copyright 2019 Nick White. -// Use of this source code is governed by the GPLv3 -// license that can be found in the LICENSE file. - -package main - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/preproc" -) - -// TODO: do more testing to see how good this assumption is -func autowsize(bounds image.Rectangle) int { - return bounds.Dx() / 60 -} - -func main() { - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: binarize [-k num] [-t type] [-w num] inimg outimg\n") - flag.PrintDefaults() - } - wsize := flag.Int("w", 0, "Window size for sauvola algorithm. Set automatically based on resolution if not set.") - ksize := flag.Float64("k", 0.5, "K for sauvola algorithm. This controls the overall threshold level. Set it lower for very light text (try 0.1 or 0.2).") - btype := flag.String("t", "binary", "Type of threshold. binary or zeroinv are currently implemented.") - 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) - - if *wsize == 0 { - *wsize = autowsize(b) - log.Printf("Set window size to %d\n", *wsize) - } - - if *wsize%2 == 0 { - *wsize++ - } - - // TODO: come up with a way to set a good ksize automatically - - var thresh image.Image - thresh = preproc.IntegralSauvola(gray, *ksize, *wsize) - - if *btype == "zeroinv" { - thresh, err = preproc.BinToZeroInv(thresh.(*image.Gray), img.(*image.RGBA)) - if err != nil { - log.Fatal(err) - } - } - - f, err = os.Create(flag.Arg(1)) - if err != nil { - log.Fatalf("Could not create file %s: %v\n", flag.Arg(1), err) - } - defer f.Close() - err = png.Encode(f, thresh) - if err != nil { - log.Fatalf("Could not encode image: %v\n", err) - } -} diff --git a/cmd/preproc/main.go b/cmd/preproc/main.go deleted file mode 100644 index 3323635..0000000 --- a/cmd/preproc/main.go +++ /dev/null @@ -1,94 +0,0 @@ -// Copyright 2019 Nick White. -// Use of this source code is governed by the GPLv3 -// license that can be found in the LICENSE file. - -package main - -// TODO: come up with a way to set a good ksize automatically - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/preproc" -) - -// TODO: do more testing to see how good this assumption is -func autowsize(bounds image.Rectangle) int { - return bounds.Dx() / 60 -} - -func main() { - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: preproc [-bt bintype] [-bw winsize] [-k num] [-m minperc] [-nowipe] [-wt wipethresh] [-ws wipesize] inimg outimg\n") - fmt.Fprintf(os.Stderr, "Binarize and preprocess an image\n") - flag.PrintDefaults() - } - binwsize := flag.Int("bw", 0, "Window size for sauvola binarization algorithm. Set automatically based on resolution if not set.") - ksize := flag.Float64("k", 0.5, "K for sauvola binarization algorithm. This controls the overall threshold level. Set it lower for very light text (try 0.1 or 0.2).") - btype := flag.String("bt", "binary", "Type of binarization threshold. binary or zeroinv are currently implemented.") - min := flag.Int("m", 30, "Minimum percentage of the image width for the content width calculation to be considered valid.") - nowipe := flag.Bool("nowipe", false, "Disable wiping completely.") - wipewsize := flag.Int("ws", 5, "Window size for wiping algorithm.") - thresh := flag.Float64("wt", 0.05, "Threshold for the wiping algorithm to determine the proportion of black pixels below which a window is determined to be the edge.") - 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) - - if *binwsize == 0 { - *binwsize = autowsize(b) - } - - if *binwsize%2 == 0 { - *binwsize++ - } - - log.Print("Binarising") - var clean, threshimg image.Image - threshimg = preproc.IntegralSauvola(gray, *ksize, *binwsize) - - if *btype == "zeroinv" { - threshimg, err = preproc.BinToZeroInv(threshimg.(*image.Gray), img.(*image.RGBA)) - if err != nil { - log.Fatal(err) - } - } - - if !*nowipe { - log.Print("Wiping sides") - clean = preproc.Wipe(threshimg.(*image.Gray), *wipewsize, *thresh, *min) - } else { - clean = threshimg - } - - f, err = os.Create(flag.Arg(1)) - if err != nil { - log.Fatalf("Could not create file %s: %v\n", flag.Arg(1), err) - } - defer f.Close() - err = png.Encode(f, clean) - if err != nil { - log.Fatalf("Could not encode image: %v\n", err) - } -} diff --git a/cmd/preprocmulti/main.go b/cmd/preprocmulti/main.go deleted file mode 100644 index aeac767..0000000 --- a/cmd/preprocmulti/main.go +++ /dev/null @@ -1,105 +0,0 @@ -// Copyright 2019 Nick White. -// Use of this source code is governed by the GPLv3 -// license that can be found in the LICENSE file. - -package main - -// TODO: come up with a way to set a good ksize automatically - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/preproc" - "rescribe.xyz/preproc/integralimg" -) - -// TODO: do more testing to see how good this assumption is -func autowsize(bounds image.Rectangle) int { - return bounds.Dx() / 60 -} - -func main() { - ksizes := []float64{0.1, 0.2, 0.4, 0.5} - - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: preprocmulti [-bt bintype] [-bw winsize] [-m minperc] [-nowipe] [-ws wipesize] inimg outbase\n") - fmt.Fprintf(os.Stderr, "Binarize and preprocess an image, with multiple binarisation levels,\n") - fmt.Fprintf(os.Stderr, "saving images to outbase_bin{k}.png.\n") - fmt.Fprintf(os.Stderr, "Binarises with these levels for k: %v.\n", ksizes) - flag.PrintDefaults() - } - binwsize := flag.Int("bw", 0, "Window size for sauvola binarization algorithm. Set automatically based on resolution if not set.") - btype := flag.String("bt", "binary", "Type of binarization threshold. binary or zeroinv are currently implemented.") - min := flag.Int("m", 30, "Minimum percentage of the image width for the content width calculation to be considered valid.") - nowipe := flag.Bool("nowipe", false, "Disable wiping completely.") - wipewsize := flag.Int("ws", 5, "Window size for wiping algorithm.") - flag.Parse() - if flag.NArg() < 2 { - flag.Usage() - os.Exit(1) - } - - log.Printf("Opening %s\n", flag.Arg(0)) - 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) - - if *binwsize == 0 { - *binwsize = autowsize(b) - } - - if *binwsize%2 == 0 { - *binwsize++ - } - - var clean, threshimg image.Image - log.Print("Precalculating integral images") - integrals := integralimg.ToAllIntegralImg(gray) - - for _, k := range ksizes { - log.Print("Binarising") - threshimg = preproc.PreCalcedSauvola(integrals, gray, k, *binwsize) - - if *btype == "zeroinv" { - threshimg, err = preproc.BinToZeroInv(threshimg.(*image.Gray), img.(*image.RGBA)) - if err != nil { - log.Fatal(err) - } - } - - if !*nowipe { - log.Print("Wiping sides") - clean = preproc.Wipe(threshimg.(*image.Gray), *wipewsize, k*0.02, *min) - } else { - clean = threshimg - } - - savefn := fmt.Sprintf("%s_bin%0.1f.png", flag.Arg(1), k) - log.Printf("Saving %s\n", savefn) - f, err = os.Create(savefn) - if err != nil { - log.Fatalf("Could not create file %s: %v\n", savefn, err) - } - defer f.Close() - err = png.Encode(f, clean) - if err != nil { - log.Fatalf("Could not encode image: %v\n", err) - } - } -} diff --git a/cmd/splittable/main.go b/cmd/splittable/main.go deleted file mode 100644 index e6df94a..0000000 --- a/cmd/splittable/main.go +++ /dev/null @@ -1,143 +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. - -package main - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/preproc/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.I, 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.I, 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.I, wsize int, thresh float64) []int { - maxx := len(img[0]) - 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.ToIntegralImg(gray) - 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 -} diff --git a/cmd/wipe/main.go b/cmd/wipe/main.go deleted file mode 100644 index d4c95ac..0000000 --- a/cmd/wipe/main.go +++ /dev/null @@ -1,59 +0,0 @@ -// Copyright 2019 Nick White. -// Use of this source code is governed by the GPLv3 -// license that can be found in the LICENSE file. - -package main - -import ( - "flag" - "fmt" - "image" - "image/draw" - _ "image/jpeg" - "image/png" - "log" - "os" - - "rescribe.xyz/preproc" -) - -func main() { - flag.Usage = func() { - fmt.Fprintf(os.Stderr, "Usage: wipe [-m minperc] [-t thresh] [-w winsize] inimg outimg\n") - fmt.Fprintf(os.Stderr, "Wipes the sections of an image which are outside the content area.\n") - flag.PrintDefaults() - } - min := flag.Int("m", 30, "Minimum percentage of the image width for the content width calculation to be considered valid.") - thresh := flag.Float64("t", 0.05, "Threshold for the proportion of black pixels below which a window is determined to be the edge. Higher means more aggressive wiping.") - wsize := flag.Int("w", 5, "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) - - clean := preproc.Wipe(gray, *wsize, *thresh, *min) - - f, err = os.Create(flag.Arg(1)) - if err != nil { - log.Fatalf("Could not create file %s: %v\n", flag.Arg(1), err) - } - defer f.Close() - err = png.Encode(f, clean) - if err != nil { - log.Fatalf("Could not encode image: %v\n", err) - } -} |