Noticed that images sometimes are sliced up in PDFs.
Steps:
insert an image with a high resoultion (3000x1800) into a .docx
use "Microsoft Print to PDF" option of Word to convert to PDF
extracting all images with pdfimages or pymupdf
Result:
Image is sliced horizontally into three images
Questions:
What exactly happens in the in the transition from .docx to pdf (or in generell in the process to pdf) that makes the converter slice it up into three images instead of one?
Do the individuell XObjects of the sliced images contain information which says that these three images belong to originally one?
How do I know how the images are sliced (horizontally / vertically) and what if originally there were two images inserted into the .docx file and both of them are sliced. Can you tell if slice x belongs to original image y or z?
So, as you have found out: because the code which generates the PDF choose to do so.
The technical reasons may be various - it could be that historically there were printers which would only have so much memory, and would need to get limiterd size-images when printing, and someone at some point when writing the PDF export code present in Microsoft Office choose to apply this limit.
Anyway, technically, as put in the comments, an image in a PDF file could be composed of unlimited smaller images collated together.
Now, the second part, and your actual question: to know whether images ibn a PDF file belong together in a single original image one would need a custom extractor tool to check the geometry of all images in the document and find out which images have no margins or boundaries with others - it would not be that hard to do for well behaved files (which we can't know if MS Office generated files are: there are ways to obfuscate image positioning by making it indirectly). The metadata in the image-parts may or may not contain information that would allow one to recompose the original image: it would be up to the code generating the PDF to include this metadata or not - but the geometry can't lie in this case: if the final document presents a single image visually, it is possible to detect that when fetching the images.
Related
The question seems to be weird, but I need to ask this, since I am witnessing a quite interesting output when I compare text as image and graphics as image.
Ideally I am in process of identifying an tool, or algorithm to compare two pdfs, generate output which will highlight the difference between them.
There are possibilities in pdfs, which will have text as image format (legacy text on papers, are converted to pdfs).
and we are doing migration of those legacy pdfs, and finally we are comparing with legacy and converted pdf output.
I am evaluating couple of tools like Adobe dc pro, i-net pdfc and power pdf etc, for comparing two pdfs.
While evaluating, I am able to see graphic images are getting compared(not accurate either) on either side of the pdfs. Where as text as images are completely ignored, unanimously same results in all the tools.
But I am more interested in text as image, since we deal more of legacy text pdfs.
Below, is attached graphic image comparison result, where it could able to capture the differences between the images.
But when I compare text image, differences are not highlighted in the tool.
What I understand from this, text is not compared as image graphics, and tool is completely ignoring the comparison. I would like have clarification whether my assumption is correct.
Secondly, I would like to know how to compare text image in pdfs to generate the differences?.
I'm working for the company that is author of i-net PDFC so I'll answer your first question as well:
Your assumption is correct. i-net PDFC is able to compare images and shapes, but it cannot detect if some content completely changed it's meaning, e.G. a line shape that is used to draw a letter or in your case an image that has to be recognized as text. Recognizing ASCII art as image won't work for the same reason either. Such cases will always be detected as differences even though their visual appearance is similar.
On your second question: Using an OCR conversion tool for one or both documents is a common solution to this problem. A simple image comparison of the compared pages in unlikely to work due to the different font styles and line wrappings in the converted file.
Please note that most OCR applications will use the rendered page images for the recognition. This may lead to incorrect recognition results even if there are no images in the PDF file.
i-net Software is aware of this general issue and an OCR module is currently in development. It'll provide an option to apply the recognition solely to the images in the PDF files.
I'm currently trying to achieve this:
I have a very large TIFF image, which contains scanned documents. The image contains invoices with barcodes/QR codes, followed by multiple other scanned documents related to the invoice which preceded them. This can be repeated multiple times ( the TIFF image may look like [invoice] + [documents] + [invoice] + [documents] ... )
I need a program (doesn't really matter in which language but I'd prefer either Java, JavaScript, PHP, C++ or Python) that takes said TIFF image, scans all the barcodes and returns their values and their position in the image (either which page it is on or it's absolute position, but the page is preferable, I know for certain that there won't be multiple barcodes on one page). The goal is to split this TIFF image into multiple PDF files, each containing only one invoice and all of the documents that belong to the invoice.
I have the latter part done already. I intend to use ImageMagick to split the TIFF file into multiple files (tested, works). I have also tried multiple barcode scanning methods, but met critical problems at every one. And that's the point of my question:
Is any of my presumptions false? Is there a better way/library/SW that you know about that could work?
Libraries/SW I tried so far:
ZXing port for PHP: Can't work with TIFF files
ZXing github
Quagga for JavaScript: Can't work with TIFF files either.
Quagga github
ZBar code reader: The best looking one by far. I managed to scan multiple QR codes in one TIFF image using CMD (Windows), but didn't find a way to get their positions. Also found out that C++ and Python versions exist, but didn't get to try them out just yet.
Thanks for any ideas/corrections.
The best one I heard -that is subjective ofc- is Barcode Rendering Framework
I'm not sure if it can detect multiple barcodes on a page but it can detect many different types of barcodes.
And it's also Open Source..
I have (or, rather, will soon have) a number of maps created in ArcGIS 10.0 and exported as PDF documents. The maps all show contiguous areas, being rather like the pages in a map book. There will also be a smaller-scale map depicting the entire area (let's call it the "study area"), but with less detail, rather like that page of a map atlas that shows what page depicts what area.
I wonder if there is any way to create thumbnails of the larger-scale maps and mosaic them such as to create an index map of the study area. A user would then be able to see, for a particular point on the smaller-scale map, which of the larger-scale maps depicts that part of the study area. (And perhaps see that map by clicking on the larger map?) Does anyone have any ideas I can implement this? I would prefer exporting the maps in PDF format, but, if I can't do all of the above with PDF, then any other format to which a map can be exported from ArcGIS, such as JPG or TIF, will work.
You should be able to create a PDF which does this.
What you need to do is render each page to a small image.
Then collect each of these images and add them as a mosaic to an index page.
Then put links from each small image back to the original PDF page.
If the hierarchy was more than one level deep you could repeat the process.
You need a PDF component to do this. What you want in terms of features is something which does decent PDF rendering. It's an easy thing to do badly and a difficult thing to do well.
ABCpdf .NET does good quality rendering so it's what I would suggest, but then I would because I work on it. :-)
I've read much about PDF extractions and libraries (as iText) but i just haven't found a solution to extract images and text (with coordinates) from a PDF.
The task is to scan PDF with catalog of products and extract each image. There is an image code printed next to each image and also a list of product codes for products that are shown on the image.
I know that there is no way to extract structured info from a PDF like this but with coordinates of all image and text objects I could write code to identify linked text by its distance from the image. Then I could split text using a RegExp and find out what is a product code, what is an image code etc.
Could you recommend a good and working solution for the task?
Use XPDF (http://www.foolabs.com/xpdf/)
It can extract all the characters in the PDF with co-ordinates (pdftotext -bbox [sourcefile] [outputfile]) and also all the images and SVGs in the PDF.
It's open source (GPLv2) and supports a lot of additional extraction functionalities as well.
Several Java libraries can do this. Have you looked at JPedal or PdfBox?
If a commercial library is an option for you, you could try Amyuni PDF Creator .Net or Amyuni PDF Creator ActiveX. You could use the method IacDocument.GetObjectsInRectangle to retrieve all the "graphic objects" of your interest, then use the ObjectType attribute to separate images from text. The library already provides an algorithm for putting close text together. From the documentation:
IacDocument.GetObjectsInRectangle Method
The GetObjectsInRectangle method gets all the objects that are in the specified rectangle.
Usual disclaimer applies.
I ran into this trying to throw together a simple Automator script to combine several one-page PDF files. I had 88 files to combine, each just about exactly 300KB, so I expected the final product to be about 30MB; the resulting PDF file, using the Combine PDFs Automator action, was 300+MB.
Poking around, the Automator action uses a Python script, with Foundation bindings, to create the new PDF document with the CoreGraphics PDF APIs. Nothing seems out of place. Basically, it's doing this (simplified, but these are the high points):
writeContext = CGPDFContextCreateWithURL(outURL, None, None)
for url in inURLs:
doc = CGPDFDocumentCreateWithURL(url)
page = CGPDFDocumentGetPage(doc, 1)
mediaBox = CGPDFPageGetBoxRect(page, kCGPDFMediaBox)
CGContextBeginPage(writeContext, mediaBox)
CGContextDrawPDFPage(writeContext, page)
CGContextEndPage(writeContext)
CGPDFContextClose(writeContext)
I can't imagine that CGContextDrawPDFPage, when drawing to a PDF context, would do anything but copy the PDF data for that page (with some window-dressing).
Even when "combining" just one PDF, the output is 2.8MB, compared to the 300KB original one-page PDF.
The resulting PDFs look exactly the same page-by-page as the original pages: text is selectable in the same places, graphics look identical, the pages are exactly the same size.
Any ideas?
Do the input PDFs contain the same set of fonts, or different sets? Maybe if the originals don't contain embedded fonts, but the output does, that could account for some of the growth.