mirror of
https://github.com/unidoc/unipdf.git
synced 2025-05-01 22:17:29 +08:00

* Prepared skeleton and basic component implementations for the jbig2 encoding. * Added Bitset. Implemented Bitmap. * Decoder with old Arithmetic Decoder * Partly working arithmetic * Working arithmetic decoder. * MMR patched. * rebuild to apache. * Working generic * Working generic * Decoded full document * Update Jenkinsfile go version [master] (#398) * Update Jenkinsfile go version * Decoded AnnexH document * Minor issues fixed. * Update README.md * Fixed generic region errors. Added benchmark. Added bitmap unpadder. Added Bitmap toImage method. * Fixed endofpage error * Added integration test. * Decoded all test files without errors. Implemented JBIG2Global. * Merged with v3 version * Fixed the EOF in the globals issue * Fixed the JBIG2 ChocolateData Decode * JBIG2 Added license information * Minor fix in jbig2 encoding. * Applied the logging convention * Cleaned unnecessary imports * Go modules clear unused imports * checked out the README.md * Moved trace to Debug. Fixed the build integrate tag in the document_decode_test.go * Initial encoder skeleton * Applied UniPDF Developer Guide. Fixed lint issues. * Cleared documentation, fixed style issues. * Added jbig2 doc.go files. Applied unipdf guide style. * Minor code style changes. * Minor naming and style issues fixes. * Minor naming changes. Style issues fixed. * Review r11 fixes. * Added JBIG2 Encoder skeleton. * Moved Document and Page to jbig2/document package. Created decoder package responsible for decoding jbig2 stream. * Implemented raster functions. * Added raster uni low test funcitons. * Added raster low test functions * untracked files on jbig2-encoder: c869089 Added raster low test functions * index on jbig2-encoder: c869089 Added raster low test functions * Added morph files. * implemented jbig2 encoder basics * JBIG2 Encoder - Generic method * Added jbig2 image encode ttests, black/white image tests * cleaned and tested jbig2 package * unfinished jbig2 classified encoder * jbig2 minor style changes * minor jbig2 encoder changes * prepared JBIG2 Encoder * Style and lint fixes * Minor changes and lints * Fixed shift unsinged value build errors * Minor naming change * Added jbig2 encode, image gondels. Fixed jbig2 decode bug. * Provided jbig2 core.DecodeGlobals function. * Fixed JBIG2Encoder `r6` revision issues. * Removed public JBIG2Encoder document. * Minor style changes * added NewJBIG2Encoder function. * fixed JBIG2Encoder 'r9' revision issues. * Cleared 'r9' commented code. * Updated ACKNOWLEDGEMENETS. Fixed JBIG2Encoder 'r10' revision issues. Co-authored-by: Gunnsteinn Hall <gunnsteinn.hall@gmail.com>
279 lines
8.3 KiB
Go
279 lines
8.3 KiB
Go
/*
|
|
* This file is subject to the terms and conditions defined in
|
|
* file 'LICENSE.md', which is part of this source code package.
|
|
*/
|
|
|
|
package classer
|
|
|
|
import (
|
|
"github.com/unidoc/unipdf/v3/common"
|
|
|
|
"github.com/unidoc/unipdf/v3/internal/jbig2/basic"
|
|
"github.com/unidoc/unipdf/v3/internal/jbig2/bitmap"
|
|
"github.com/unidoc/unipdf/v3/internal/jbig2/errors"
|
|
)
|
|
|
|
// classifyRankHaus is the classification using windowed rank hausdorff metric.
|
|
func (c *Classer) classifyRankHaus(boxa *bitmap.Boxes, pixa *bitmap.Bitmaps, pageNumber int) error {
|
|
const processName = "classifyRankHaus"
|
|
if boxa == nil {
|
|
return errors.Error(processName, "boxa not defined")
|
|
}
|
|
if pixa == nil {
|
|
return errors.Error(processName, "pixa not defined")
|
|
}
|
|
|
|
n := len(pixa.Values)
|
|
if n == 0 {
|
|
return errors.Error(processName, "empty new components")
|
|
}
|
|
nafg := pixa.CountPixels()
|
|
|
|
size := c.Settings.SizeHaus
|
|
sel := bitmap.SelCreateBrick(size, size, size/2, size/2, bitmap.SelHit)
|
|
bms1 := &bitmap.Bitmaps{Values: make([]*bitmap.Bitmap, n)}
|
|
bms2 := &bitmap.Bitmaps{Values: make([]*bitmap.Bitmap, n)}
|
|
var (
|
|
bm, bm1, bm2 *bitmap.Bitmap
|
|
err error
|
|
)
|
|
for i := 0; i < n; i++ {
|
|
bm, err = pixa.GetBitmap(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
|
|
bm1, err = bm.AddBorderGeneral(JbAddedPixels, JbAddedPixels, JbAddedPixels, JbAddedPixels, 0)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
bm2, err = bitmap.Dilate(nil, bm1, sel)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
bms1.Values[n] = bm1 // un-dilated
|
|
bms2.Values[n] = bm2 // dilated
|
|
}
|
|
pta, err := bitmap.Centroids(bms1.Values)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
if err = pta.Add(c.CentroidPoints); err != nil {
|
|
common.Log.Trace("No centroids to add")
|
|
}
|
|
|
|
if c.Settings.RankHaus == 1.0 {
|
|
err = c.classifyRankHouseOne(boxa, pixa, bms1, bms2, pta, pageNumber)
|
|
} else {
|
|
err = c.classifyRankHouseNonOne(boxa, pixa, bms1, bms2, pta, nafg, pageNumber)
|
|
}
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *Classer) classifyRankHouseOne(boxa *bitmap.Boxes, pixa, bms1, bms2 *bitmap.Bitmaps, pta *bitmap.Points, pageNumber int) (err error) {
|
|
const processName = "Classer.classifyRankHouseOne"
|
|
var (
|
|
x1, y1, x2, y2 float32
|
|
iClass int
|
|
bm, bm1, bm2, bm3, bm4 *bitmap.Bitmap
|
|
found, testVal bool
|
|
)
|
|
for i := 0; i < len(pixa.Values); i++ {
|
|
bm1 = bms1.Values[i]
|
|
bm2 = bms2.Values[i]
|
|
x1, y1, err = pta.GetGeometry(i)
|
|
if err != nil {
|
|
return errors.Wrapf(err, processName, "first geometry")
|
|
}
|
|
|
|
nt := len(c.UndilatedTemplates.Values)
|
|
|
|
found = false
|
|
findContext := initSimilarTemplatesFinder(c, bm1)
|
|
for iClass = findContext.Next(); iClass > -1; {
|
|
bm3, err = c.UndilatedTemplates.GetBitmap(iClass)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "bm3")
|
|
}
|
|
bm4, err = c.DilatedTemplates.GetBitmap(iClass)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "bm4")
|
|
}
|
|
|
|
x2, y2, err = c.CentroidPointsTemplates.GetGeometry(iClass)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "CentroidTemplates")
|
|
}
|
|
testVal, err = bitmap.HausTest(bm1, bm2, bm3, bm4, x1-x2, y1-y2, MaxDiffWidth, MaxDiffHeight)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
|
|
if testVal {
|
|
found = true
|
|
// add the class index in the slice
|
|
if err = c.ClassIDs.Add(iClass); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
// add the page number for the class index.
|
|
if err = c.ComponentPageNumbers.Add(pageNumber); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
|
|
if c.Settings.KeepClassInstances {
|
|
bitmaps, err := c.ClassInstances.GetBitmaps(iClass)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "KeepPixaa")
|
|
}
|
|
bm, err = pixa.GetBitmap(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "KeepPixaa")
|
|
}
|
|
bitmaps.AddBitmap(bm)
|
|
|
|
box, err := boxa.Get(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "KeepPixaa")
|
|
}
|
|
bitmaps.AddBox(box)
|
|
}
|
|
break
|
|
}
|
|
}
|
|
|
|
if !found {
|
|
if err = c.ClassIDs.Add(nt); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
if err = c.ComponentPageNumbers.Add(pageNumber); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
bitmaps := &bitmap.Bitmaps{}
|
|
bm, err = pixa.GetBitmap(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "!found")
|
|
}
|
|
bitmaps.Values = append(bitmaps.Values, bm)
|
|
wt, ht := bm.Width, bm.Height
|
|
c.TemplatesSize.Add(uint64(ht)*uint64(wt), nt)
|
|
box, err := boxa.Get(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "!found")
|
|
}
|
|
bitmaps.AddBox(box)
|
|
c.ClassInstances.AddBitmaps(bitmaps)
|
|
c.CentroidPointsTemplates.AddPoint(x1, y1)
|
|
c.UndilatedTemplates.AddBitmap(bm1)
|
|
c.DilatedTemplates.AddBitmap(bm2)
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// classifyRankHouseNonOne is a helper that classifies when the rank < 1.0.
|
|
func (c *Classer) classifyRankHouseNonOne(boxa *bitmap.Boxes, pixa, bms1, bms2 *bitmap.Bitmaps, pta *bitmap.Points, nafg *basic.NumSlice, pageNumber int) (err error) {
|
|
const processName = "Classer.classifyRankHouseOne"
|
|
var (
|
|
x1, y1, x2, y2 float32
|
|
area1, area3, iClass int
|
|
bm, bm1, bm2, bm3, bm4 *bitmap.Bitmap
|
|
found, testVal bool
|
|
)
|
|
tab8 := bitmap.MakePixelSumTab8()
|
|
for i := 0; i < len(pixa.Values); i++ {
|
|
if bm1, err = bms1.GetBitmap(i); err != nil {
|
|
return errors.Wrap(err, processName, "bms1.Get(i)")
|
|
}
|
|
|
|
if area1, err = nafg.GetInt(i); err != nil {
|
|
common.Log.Trace("Getting FGTemplates at: %d failed: %v", i, err)
|
|
}
|
|
|
|
if bm2, err = bms2.GetBitmap(i); err != nil {
|
|
return errors.Wrap(err, processName, "bms2.Get(i)")
|
|
}
|
|
if x1, y1, err = pta.GetGeometry(i); err != nil {
|
|
return errors.Wrapf(err, processName, "pta[i].Geometry")
|
|
}
|
|
|
|
nt := len(c.UndilatedTemplates.Values)
|
|
found = false
|
|
findContext := initSimilarTemplatesFinder(c, bm1)
|
|
for iClass = findContext.Next(); iClass > -1; {
|
|
if bm3, err = c.UndilatedTemplates.GetBitmap(iClass); err != nil {
|
|
return errors.Wrap(err, processName, "pixat.[iClass]")
|
|
}
|
|
if area3, err = c.FgTemplates.GetInt(iClass); err != nil {
|
|
common.Log.Trace("Getting FGTemplate[%d] failed: %v", iClass, err)
|
|
}
|
|
|
|
if bm4, err = c.DilatedTemplates.GetBitmap(iClass); err != nil {
|
|
return errors.Wrap(err, processName, "pixatd[iClass]")
|
|
}
|
|
if x2, y2, err = c.CentroidPointsTemplates.GetGeometry(iClass); err != nil {
|
|
return errors.Wrap(err, processName, "CentroidPointsTemplates[iClass]")
|
|
}
|
|
testVal, err = bitmap.RankHausTest(bm1, bm2, bm3, bm4, x1-x2, y1-y2, MaxDiffWidth, MaxDiffHeight, area1, area3, float32(c.Settings.RankHaus), tab8)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
|
|
if testVal {
|
|
found = true
|
|
if err = c.ClassIDs.Add(iClass); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
if err = c.ComponentPageNumbers.Add(pageNumber); err != nil {
|
|
return errors.Wrap(err, processName, "")
|
|
}
|
|
if c.Settings.KeepClassInstances {
|
|
bitmaps, err := c.ClassInstances.GetBitmaps(iClass)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "c.Pixaa.GetBitmaps(iClass)")
|
|
}
|
|
if bm, err = pixa.GetBitmap(i); err != nil {
|
|
return errors.Wrap(err, processName, "pixa[i]")
|
|
}
|
|
bitmaps.Values = append(bitmaps.Values, bm)
|
|
box, err := boxa.Get(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "boxa.Get(i)")
|
|
}
|
|
bitmaps.Boxes = append(bitmaps.Boxes, box)
|
|
}
|
|
break
|
|
}
|
|
}
|
|
if !found {
|
|
if err = c.ClassIDs.Add(nt); err != nil {
|
|
return errors.Wrap(err, processName, "!found")
|
|
}
|
|
if err = c.ComponentPageNumbers.Add(pageNumber); err != nil {
|
|
return errors.Wrap(err, processName, "!found")
|
|
}
|
|
|
|
bitmaps := &bitmap.Bitmaps{}
|
|
bm = pixa.Values[i]
|
|
bitmaps.AddBitmap(bm)
|
|
|
|
wt, ht := bm.Width, bm.Height
|
|
c.TemplatesSize.Add(uint64(wt)*uint64(ht), nt)
|
|
box, err := boxa.Get(i)
|
|
if err != nil {
|
|
return errors.Wrap(err, processName, "!found")
|
|
}
|
|
bitmaps.AddBox(box)
|
|
c.ClassInstances.AddBitmaps(bitmaps)
|
|
c.CentroidPointsTemplates.AddPoint(x1, y1)
|
|
c.UndilatedTemplates.AddBitmap(bm1)
|
|
c.DilatedTemplates.AddBitmap(bm2)
|
|
c.FgTemplates.AddInt(area1)
|
|
}
|
|
}
|
|
c.NumberOfClasses = len(c.UndilatedTemplates.Values)
|
|
return nil
|
|
}
|