internal static void ProcessWords(Bitmap source, int screen, string text, Point location, PageSections pageSections)
{
List<List<Rectangle>> imageWords = new List<List<Rectangle>>();
string[] textWords = text.Split(new char[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
List<Rectangle> letters = new List<Rectangle>();
bool active = false;
bool inDescription = pageSections.DescriptiveText.Bounds.Contains(location);
for (int j = 0, i = 0; i < pageSections.AllLetters.Count; i++)
{
if (pageSections.AllLetters[i].Contains(location))
{
active = true;
}
if (active)
{
if (j > 0 && ImageLetters.IsNewWord(pageSections.AllLetters, i, inDescription))
{
imageWords.Add(letters);
letters = new List<Rectangle>();
}
letters.Add(pageSections.AllLetters[i]);
j++;
}
}
imageWords.Add(letters);
for (int i = 0; i < Math.Min(imageWords.Count, textWords.Length); i++)
{
// Test if words in text and gaps in bitmap text match up.
// We need this to reject any text where letters were not isolated properly.
if (imageWords[i].Count != textWords[i].Length)
{
continue;
}
for (int j = 0; j < textWords[i].Length; j++)
{
TeachLetters.ProcessLetter(source, screen, textWords[i][j], imageWords[i][j]);
}
}
return;
}
private void PrepareBitmaps(Bitmap bmp)
{
DisposeOldImages();
Bitmap binary, binarySplit;
pageSections = LibExplOCR.PrepareBitmaps(bmp, out baseBmp, out binary, out binarySplit);
pageSections = ImageLetters.RefinePartition(pageSections, binary);
drawBmpRaw = new Bitmap(binary);
drawBmpRawSplit = new Bitmap(binarySplit);
drawBmp = new Bitmap(binary);
LibExplOCR.AnnotatePageStructure(drawBmp, pageSections);
binary.Dispose();
}
internal static void AnnotatePageStructure(Bitmap bmp, PageSections sections)
{
if (sections == null)
{
return;
}
using (Graphics g = Graphics.FromImage(bmp))
using (Pen greenpen = new Pen(Color.FromArgb(120, Color.Green)))
using (Brush brushHatch = new HatchBrush(HatchStyle.BackwardDiagonal, Color.Purple))
using (Brush b = new SolidBrush(Color.FromArgb(120, Color.Blue)))
{
foreach (ExcludeSection exclude in sections.Excluded)
{
g.FillRectangle(brushHatch, exclude.Bounds);
g.DrawRectangle(Pens.Purple, exclude.Bounds);
}
foreach (Line line in sections.AllLines)
{
g.FillRectangle(b, line.Bounds);
foreach (Rectangle letter in line)
{
g.DrawRectangle(greenpen, letter);
}
}
if (sections.DescriptiveText != null)
{
Rectangle r = sections.DescriptiveText.Bounds;
r.Inflate(-3, 0);
g.DrawRectangle(Pens.Orange, r);
}
foreach (TextLineSection tl in sections.TextLines)
{
g.DrawRectangle(Pens.Purple, tl.Line.Bounds);
}
foreach (HeadlineSection hl in sections.Headlines)
{
g.DrawRectangle(Pens.Khaki, hl.Line.Bounds);
}
foreach (TableSection table in sections.Tables)
{
g.DrawRectangle(Pens.Red, table.Bounds);
g.DrawRectangle(Pens.Red, table.Gap);
int previousItem = 0;
for (int i = 1 /*sic!*/; i < table.Count; i++)
{
if (table.GetLineItem(i) > previousItem)
{
g.DrawLine(Pens.Red, table.Bounds.Left, table[i].Bounds.Top - 6, table.Gap.Left, table[i].Bounds.Top - 6);
g.DrawLine(Pens.Red, table.Bounds.Right, table[i].Bounds.Top - 6, table.Gap.Right, table[i].Bounds.Top - 6);
previousItem = table.GetLineItem(i);
}
}
}
/* Word display for debugging.
int wordStart = 0;
for (int i = 0; i < sections.AllLetters.Count; i++)
{
bool small = sections.DescriptiveText.Bounds.Contains(sections.AllLetters[i]);
if (ImageLetters.IsNewWord(sections.AllLetters,i, small))
{
r = sections.AllLetters[wordStart];
for (int j = wordStart; j < i; j++)
{
r = Rectangle.Union(r, sections.AllLetters[j]);
}
g.FillRectangle(Brushes.Yellow, r);
wordStart = i;
}
}*/
}
}
internal static PageSections RefinePartition(PageSections pageSections, Bitmap binary)
{
Bytemap imageBinary = new Bytemap(binary);
List<Line> descriptionLines = new List<Line>();
if (pageSections.DescriptiveText == null)
{
return pageSections;
}
foreach (Line line in pageSections.DescriptiveText)
{
List<Rectangle> accumulate = new List<Rectangle>();
foreach (Rectangle letter in line)
{
Bytemap letterMask = ImageLetters.CopyRectangle(imageBinary, letter);
accumulate.AddRange(ImageLetters.CleanupKerning(letterMask, false));
}
descriptionLines.Add(new Line(line.Bounds, accumulate));
}
TextSection descriptiveText = new TextSection(descriptionLines);
// Fix kerning for all text lines - hoping for terraforming and mining resources lines.
List<TextLineSection> textLines = new List<TextLineSection>();
foreach (TextLineSection tls in pageSections.TextLines)
{
List<Rectangle> accumulate = new List<Rectangle>();
foreach (Rectangle letter in tls.Line)
{
Bytemap letterMask = ImageLetters.CopyRectangle(imageBinary, letter);
accumulate.AddRange(ImageLetters.CleanupKerning(letterMask, false));
}
textLines.Add(new TextLineSection(new Line(tls.Line.Bounds, accumulate)));
}
return new PageSections(pageSections.Tables, descriptiveText, textLines, pageSections.Excluded, pageSections.Headlines);
}
internal void ReadPage(Bytemap imageGray, Bytemap imageBinary, Bytemap imageBinarySplit, PageSections sections)
{
int descriptionLimit = -1;
qualityData.Clear();
List<TransferItem> output = new List<TransferItem>();
// Get rid of those pesky powerplay tables.
foreach (IPageSection section in sections.AllSections)
{
if (section is TextSection)
{
descriptionLimit = section.Bounds.Bottom;
}
}
foreach (IPageSection section in sections.AllSections)
{
if (section is HeadlineSection)
{
TransferItem ti = ReadHeadline(section as HeadlineSection, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
}
if (section is TextSection)
{
output.Add(ReadDescription(section as TextSection, imageGray, imageBinary));
}
if (section is TableSection)
{
if (descriptionLimit > section.Bounds.Top)
{
continue;
}
output.AddRange(ReadTableSection(section as TableSection, sections, imageGray, imageBinary, imageBinarySplit));
}
if (section is TextLineSection)
{
TransferItem ti;
TextLineSection tsl = section as TextLineSection;
ti = ReadTerraformingLine(tsl, sections, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
ti = ReadMiningReservesLine(tsl, sections, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
}
}
CustomItemProcessing(output);
if (StitchPrevious)
{
output = MergeItems(currentItems, output);
}
AppendMetaInformation(output);
currentItems = output.ToArray();
}
private TransferItem ReadTerraformingLine(TextLineSection tsl, PageSections sections, Bytemap imageGray, Bytemap imageBinary)
{
// Terraforming description is above the first table.
if (sections.Tables.Count < 1) return null;
if (tsl.Line.Bounds.Bottom >= sections.Tables[0].Bounds.Top) return null;
string terraforming = "";
List<Rectangle> rs = new List<Rectangle>(tsl.Line);
for (int i = 0; i < tsl.Line.Count; i++)
{
if (i > 0 && ImageLetters.IsNewWord(rs, i, true))
{
terraforming += " ";
}
terraforming += PredictAsLetterD(imageGray, imageBinary, rs[i]);
}
return GuessTerraforming(terraforming);
}
private TransferItem ReadMiningReservesLine(TextLineSection tsl, PageSections sections, Bytemap imageGray, Bytemap imageBinary)
{
bool afterTable = false;
bool afterDescription = false;
// Mining reserves are stated between first table and a headline OR immediately after the description..
if (sections.Tables.Count < 1) return null;
afterTable = tsl.Line.Bounds.Top > sections.Tables[0].Bounds.Bottom;
afterDescription = tsl.Line.Bounds.Bottom < sections.Tables[0].Bounds.Top
&& tsl.Line.Bounds.Bottom > sections.Tables[0].Bounds.Top - 50;
if (!afterTable && !afterDescription) return null;
int index = sections.AllSections.IndexOf(tsl);
if (index < 0 || index + 1 >= sections.AllSections.Count) return null;
if (!afterDescription && !(sections.AllSections[index + 1] is HeadlineSection)) return null;
string mining = "";
List<Rectangle> rs = new List<Rectangle>(tsl.Line);
for (int i = 0; i < tsl.Line.Count; i++)
{
if (i > 0 && ImageLetters.IsNewWord(rs, i, true))
{
mining += " ";
}
mining += PredictAsLetterD(imageGray, imageBinary, rs[i]);
}
return GuessMiningReserves(mining);
}
private IEnumerable<TransferItem> ReadTableSection(TableSection table, PageSections sections, Bytemap imageGray, Bytemap imageBinary, Bytemap imageBinarySplit)
{
List<TransferItem> tis = new List<TransferItem>();
// TODO: For now, do not allow table items that are above the
// description. This would be configurable in the future.
if (sections.DescriptiveText != null && table.Bounds.Bottom < sections.DescriptiveText.Bounds.Top)
{
return tis;
}
if (!HasLeftText(table) || !HasRightText(table))
{
return tis;
}
tis.Add(new TransferItem("DELIMITER"));
for (int i = 0; i < table.Count; i++)
{
List<Line> left;
List<Line> right;
GetTableItem(table, i, out left, out right);
tis.Add(ReadTableItem(imageGray, imageBinary, imageBinarySplit, left, right));
i += (left.Count - 1);
}
return tis;
}
internal void ReadPageClassic(Bytemap imageGray, Bytemap imageBinary, Bytemap imageBinarySplit, PageSections sections)
{
qualityData.Clear();
List<TransferItem> output = new List<TransferItem>();
if (sections.DescriptiveText != null)
{
output.Add(ReadDescription(sections.DescriptiveText, imageGray, imageBinary));
}
foreach (HeadlineSection hl in sections.Headlines)
{
TransferItem ti = ReadHeadline(hl, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
}
foreach (TableSection table in sections.Tables)
{
output.AddRange(ReadTableSection(table, sections, imageGray, imageBinary, imageBinarySplit));
}
foreach (TextLineSection tsl in sections.TextLines)
{
TransferItem ti;
ti = ReadTerraformingLine(tsl, sections, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
ti = ReadMiningReservesLine(tsl, sections, imageGray, imageBinary);
if (ti != null)
{
output.Add(ti);
}
}
CustomItemProcessing(output);
if (StitchPrevious)
{
output = MergeItems(currentItems, output);
}
AppendMetaInformation(output);
currentItems = output.ToArray();
}