1、识别引擎的初始化，参见上一篇文章
.NET 利用PaddleOCRSharp实现OCR文字识别
2、PDF文件转图片，使用的O2S.Components.PDFRender4NET
封装一个辅助类PdfHelper.cs
public class PdfHelper
{
/// <summary>
/// 转换的图片清晰度，1最不清晰，10最清晰
/// </summary>
public enum Definition
{
One = 1, Two = 2, Three = 3, Four = 4, Five = 5, Six = 6, Seven = 7, Eight = 8, Nine = 9, Ten = 10
}
/// <summary>
/// 将PDF文档转换为图片的方法
/// </summary>
/// <param name="pdfInputPath">PDF文件路径</param>
/// <param name="imageOutputPath">图片输出路径</param>
/// <param name="pageNum">第几页</param>
/// <param name="imageFormat">设置所需图片格式</param>
/// <param name="definition">设置图片的清晰度，数字越大越清晰</param>
/// <returns></returns>
public static (string FilePath, int MaxPage) ConvertPdf2Image(string pdfInputPath,
string imageOutputPath,
int pageNum,
ImageFormat imageFormat, Definition definition = Definition.Ten)
{
string imageName = Path.GetFileNameWithoutExtension(pdfInputPath);
PDFFile pdfFile = PDFFile.Open(pdfInputPath);
string filepath = Path.Combine(imageOutputPath, imageName + pageNum + "." + imageFormat);
if (File.Exists(filepath))
{
return (filepath, pdfFile.PageCount);
}
if (!Directory.Exists(imageOutputPath))
{
Directory.CreateDirectory(imageOutputPath);
}
if (pageNum <= 0)
{
pageNum = 1;
}
if (pageNum > pdfFile.PageCount)
{
pageNum = pdfFile.PageCount;
}
Bitmap pageImage = pdfFile.GetPageImage(pageNum - 1, 56 * (int)definition);
int canKao = pageImage.Width > pageImage.Height ? pageImage.Height : pageImage.Width;
int newHeight = canKao > 1080 ? pageImage.Height / 2 : pageImage.Height;
int newWidth = canKao > 1080 ? pageImage.Width / 2 : pageImage.Width;
Bitmap newPageImage = new Bitmap(newWidth, newHeight);
Graphics g = Graphics.FromImage(newPageImage);
g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
g.DrawImage(pageImage, new Rectangle(0, 0, newWidth, newHeight),
new Rectangle(0, 40, pageImage.Width, pageImage.Height - 40), GraphicsUnit.Pixel);
newPageImage.Save(filepath, imageFormat);
g.Dispose();
newPageImage.Dispose();
pageImage.Dispose();
pdfFile.Dispose();
return (filepath, pdfFile.PageCount);
}
/// <summary>
/// 将1个PDF文档所有页全部转换为图片
/// </summary>
/// <param name="pdfInputPath">PDF文件路径</param>
/// <param name="imageOutputPath">图片输出路径</param>
/// <param name="imageName">生成图片的名字</param>
/// <param name="imageFormat">设置所需图片格式</param>
/// <param name="definition">设置图片的清晰度，数字越大越清晰</param>
/// <param name="merge">是否合并</param>
public static void ConvertAllPdf2Images(string pdfInputPath, string imageOutputPath, string imageName, ImageFormat imageFormat, Definition definition, bool merge, decimal start, decimal end)
{
PDFFile pdfFile = PDFFile.Open(pdfInputPath);
if (!Directory.Exists(imageOutputPath))
{
Directory.CreateDirectory(imageOutputPath);
}
int startPageNum = 1;
int endPageNum = pdfFile.PageCount;
if (merge && endPageNum > 25)
{
endPageNum = 25;
}
if (end > endPageNum)
{
end = endPageNum;
}
if (start > end)
{
return;
}
endPageNum = Convert.ToInt32(end);
startPageNum = Convert.ToInt32(start);
List<Bitmap> mpas = new List<Bitmap>();
// var bitMap = new Bitmap[endPageNum];
for (int i = startPageNum; i <= endPageNum; i++)
{
try
{
Bitmap pageImage = pdfFile.GetPageImage(i - 1, 56 * (int)definition);
int canKao = pageImage.Width > pageImage.Height ? pageImage.Height : pageImage.Width;
int newHeight = canKao > 1080 ? pageImage.Height / 2 : pageImage.Height;
int newWidth = canKao > 1080 ? pageImage.Width / 2 : pageImage.Width;
Bitmap newPageImage = new Bitmap(newWidth, newHeight);
Graphics g = Graphics.FromImage(newPageImage);
g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
g.DrawImage(pageImage, new Rectangle(0, 0, newWidth, newHeight),
new Rectangle(0, 40, pageImage.Width, pageImage.Height - 40), GraphicsUnit.Pixel);
if (merge)
{
mpas.Add(newPageImage);
g.Dispose();
}
else
{
newPageImage.Save(imageOutputPath + "\\" + imageName + i + "." + imageFormat,
imageFormat);
g.Dispose();
newPageImage.Dispose();
pageImage.Dispose();
}
}
catch
{
}
finally
{
}
}
try
{
if (merge)
{
//合并图片
MergerImg(mpas, imageOutputPath, imageName, imageFormat);
}
}
catch (Exception ex)
{
}
pdfFile.Dispose();
}
/// <summary>
/// 合并图片
/// </summary>
/// <param name="maps"></param>
/// <returns></returns>
private static void MergerImg(List<Bitmap> maps, string imageOutputPath, string imageName, ImageFormat imageFormat)
{
int i = maps.Count;
if (i == 0)
throw new Exception("图片数不能够为0");
//else if (i == 1)
// return ;
//创建要显示的图片对象,根据参数的个数设置宽度
int width = maps[0].Width;
int height = 0;
maps.ForEach(x => { height += width * x.Height / x.Width; });
Bitmap backgroudImg = new Bitmap(width, height);
Graphics g = Graphics.FromImage(backgroudImg);
//清除画布,背景设置为白色
g.Clear(System.Drawing.Color.White);
int tempheight = 0;
for (int j = 0; j < i; j++)
{
int thisheight = width * maps[j].Height / maps[j].Width;
g.DrawImage(maps[j], 0, tempheight, width, thisheight);
tempheight += thisheight;
}
//保存图片
backgroudImg.Save(imageOutputPath + "\\" + imageName + "." + imageFormat, imageFormat);
g.Dispose();
return;
}
}3、选择文件，获取指定页的图片，并进行识别
var ret = PdfHelper.ConvertPdf2Image(PdfFilePath,[图片保存地址], [页码], ImageFormat.Png); OCRResult _ocrResult = OcrEngine.DetectText(ret.FilePath);
最终效果如下图
那么这个旋转是怎么实现的？
将pdf转图片之后，我们是把原图存下来的，因为转图片还是需要消耗时间的，如果图片存在就不会二次转换，
那么图片我们就可以旋转之后，覆盖原图，即不会损失大小，也不会太耦合
/// <summary>
/// 旋转
/// </summary>
/// <param name="left">是否向左旋转</param>
/// <param name="savePath">旋转后图片保存位置</param>
/// <returns></returns>
private Bitmap RotateBitmap(bool left, string savePath)
{
Bitmap srcBitmap = new Bitmap(_filepath);
int width = srcBitmap.Width;
int height = srcBitmap.Height;
Bitmap dstBitmap = new Bitmap(height, width, srcBitmap.PixelFormat);
using (Graphics graphics = Graphics.FromImage(dstBitmap))
{
graphics.TranslateTransform((float)height / 2, (float)width / 2);
graphics.RotateTransform((left ? -1 : 1) * 90);
graphics.TranslateTransform(-(float)width / 2, -(float)height / 2);
graphics.DrawImage(srcBitmap, new Point(0, 0));
}
srcBitmap?.Dispose();
if (File.Exists(savePath))
{
File.Delete(savePath);
}
dstBitmap.Save(savePath, ImageFormat.Png);
return dstBitmap;
}为啥要旋转，因为有些PDF中是扫描件，还是横版的，又不好看，又不太好识别。
旋转之后，再次调用识别，把识别结果标注在指定位置即可
川公网安备 51010702003150号
留下您的脚步
最近评论