Page by page conversion of PDF into TIFF with proper compression - optimization

Question

Problem
There are PDF documents with different type of objects inside. There are simple texts. There can be scanned images that are B&W, and also other images, that are true color. The resolution can be quite high for both (~1789X2711).
I need to convert the PDF into a set of single page TIFF files. There are quite good tools for that. For example Irfanview, ImageMagick. The problem is that I have to define a single compression type for all the pages.
Using JPG for all pages would result in loosing details for B&W images and they would be huge compared to lossless fax compression.
Using lossless fax for all would wanish colors and details of true color images.
Idea
It would be nice to examine the PDF page by page. I could check the content of the page. What kind of images are there inside, and which compression is recommanded for the particular page. I think this can be done with IText, but I don't know exactly, how it should be done. A second thing is that I want to do this analysis without fully reading the PDF file. Is it possible?
Maybe the fastest solution would be to create a list of pages for each compression type with IText analysis, and then to call Irfanview to process the choosen pages with the proper compression.
Any ideas and recommendations are welcome.
UPDATE:
I have now an answer. It does not cover all requirements, and its not freeware. Any opensource ideas? Maybe Java based solutions?

Answer 1

This can be done with DotImage DotPdf from Atalasoft (cue the obligatory "I work there and work on these products"). Here is how I would do this task in C#:
PdfImageSource source = new PdfImageSource(pdfStream);
while (source.HasMoreImages()) {
AtalaImage image = source.AcquireNext();
string fileName = GetNextTiffName();
using (FileStream outStm = new FileStream(fileName, FileMode.Create)) {
TiffEncoder encoder = new TiffEncoder();
encoder.Compression = SelectCompression(image.PixelFormat);
image.Save(outStm, encoder, null);
}
source.Release(image);
}
private TiffCompression SelectCompression(PixelFormat pf)
{
switch (pf) {
// 1 bit? use CCITT G4
case PixelFormat.Pixel1bbIndexed: return TiffCompression.Group4FaxEncoding;
// 24 bit? use JPEG
case PixelFormat.Pixel24bppBgr: return TiffCompression.JpegCompression;
// all else, Lzw
default: return TiffCompression.Lzw;
}
}
You can make SelectCompression do pretty much whatever you want. If you select an invalid compression for that pixel format, the encoder will use an appropriate lossless one in its place (for example, if you select CCITT for 24bit color, the encoder will instead use Lzw).
Our PDF decoder knows when a PDF page is just gray and returns a gray image. It does NOT do anything to get you to 1 bit (this is so antialiased text looks good), however you could threshold the gray image and look at the overall differences between it and the gray image to determine if it could go to 1 bit).
Here's how you could do a set of pages:
public void ExtractNPages(Stream pdfStream, params int[] pageIndexes)
{
PdfImageSource source = new PdfImageSource(pdfStream);
for (int i in pageIndexes) {
AtalaImage image = source[i]; // implied Acquire
string fileName = GetNextTiffName();
using (FileStream outStm = new FileStream(fileName, FileMode.Create)) {
TiffEncoder = new TiffEncoder();
encoder.Compression = SelectCompression(image.PixelFormat);
image.Save(outStm, encoder, null);
}
source.Release(image);
}
}
so now you can just do ExtractNPages(stm, 0, 2, 4, 6);