KPV: a PDF reader for Kindle, based on muPDF, GPLv3

[lesen54321]

Hello,

I've tried to photograph my problem.

I open the reader (sometimes it works, sometimes not), then I call my PDf, you can see the two columns per page. Then I call on the help and give "F", it seems only one column and the rest of the other column, it fits but not exactly on the page, then I have to scroll using the cursor down to the rest of the column to . read But I can not get the second Column, you see a little reading. The cursor scrolls down only when I go to the right of the cursor, I'm not au f is the second Column and the Umblättertasten right or left of the kindle, I'm just next to the left column of the next page, what could possibly be my fault for the? I also shift. F + clicked, but nothing happened.

Thank you. other, single-column pdfs, can be read very well.





Uploaded with ImageShack.us



Uploaded with ImageShack.us


Uploaded with ImageShack.us



Uploaded with ImageShack.us








regards

lesen54321

[geekmaster]

@lesen54321: The gray streak in the upper-left corner of the page looks like your PDF file is a collection of PHOTOGRAPHS of pages. Text on photographs cannot reflow, or have other text-based processing performed on it.

It is necessary to OCR that text in order to treat it as characters, and that may not have been done for your PDF file.

[lesen54321]

Hello,

oh, thank you. I think it was a scan of book sides, that means Photographs I think, wright?


How can I see, if it it just a scan or not, if the author has not written about thisß

Can I use AbbyFinereader10 for this or do I need a special OCR Software? I think, I have to convert the whole book to text, would it work than with the 2 column order?

regards lesen54321

[geekmaster]

Originally Posted by lesen54321

Hello,

oh, thank you. I think it was a scan of book sides, that means Photographs I think, wright?


How can I see, if it it just a scan or not, if the author has not written about thisß

Can I use AbbyFinereader10 for this or do I need a special OCR Software? I think, I have to convert the whole book to text, would it work than with the 2 column order?

regards lesen54321

Try viewing it in a PDF reader on your host PC (such as Adobe Reader) and see if you can copy/paste some text from it. If not, it is an image. It *looks* like an image, because of the streak I mentioned. With no text, there is no text processing.

Your OCR software should work if it supports your language. I do not know if that program automatically manages multiple columns, but many modern OCR programs do. Try it, or better yet, read its documentation to see how to do multi-column support. Some older programs required that you mark a reqion around each text column and OCR them separately. Luckily, things have progressed a long way since then.

[lesen54321]

Hello,

thank you, I will trie.

Kind regards
lesen54321

[dpavlin]

Originally Posted by lesen54321

Hi,

I tried the 2 colum navigation, but it did not fit into my page.
I trie to make a screen of it at the kindle with my camera.

KR lesen54321

Two column mode isn't magic. It assumes that two columns are in the middle of page.

However, you can modify page bounding box (even on scanned pages). You can check current bounding box by pressing x (part of page which is under bounding box will become inverted). Since you have scanned page, I assume that whole page will become inverted.

If that is case, press Shift+x and define new bounding boxes (first top-left corner and then bottom-right) to cover just interesting content. After that two column mode should work just fine.

FWIW, you can have different bounding boxes for odd and even pages or even manually zoom in to first column (if you have three column page) and it will still work, mostly because it's implementation is so simple.

Hope this helps.

[hawhill]

And to what geekmaster commented: While this is right in theory, in practice kindlepdfreader does *not* have a "reflowing" text-extraction mode.

Maybe in the future, but honestly I consider such modes an awful thing. PDF is not (or only secondary) structured text, it's layouted text. But then we have text extraction for PDF already, it just takes developer power to order the extracted lines words, create a (HTML?) document and pass that to the CREngine based reader.

@lesen54321: Well, I think it looks "normal" on your photos. Maybe the navigation is challenging, you should describe what you expect to work differently. But the look of the pages seems to be right.

[geekmaster]

Originally Posted by hawhill

And to what geekmaster commented: While this is right in theory, in practice kindlepdfreader does *not* have a "reflowing" text-extraction mode.

Maybe in the future, but honestly I consider such modes an awful thing. PDF is not (or only secondary) structured text, it's layouted text. But then we have text extraction for PDF already, it just takes developer power to order the extracted lines words, create a (HTML?) document and pass that to the CREngine based reader.

@lesen54321: Well, I think it looks "normal" on your photos. Maybe the navigation is challenging, you should describe what you expect to work differently. But the look of the pages seems to be right.

Unfortunately, I am not very familiar with muPDF, but I did successfully compile and run an earlier version, and it was quite nice.

After dpavlin's post, I figured from his details that you are probably just displaying a portion of the rendered image (or a portion of the original full-page image).

Sorry about the misdirection.

If you want to reflow text, perhaps calibre can do that for you. Personally, I prefer zooming on a PDF file to preserve its page layout format, but there were times with viewing PDF files on my palm pilot PDA (much smaller screen) that I really needed text reflow. I would think that is not needed with a 600x800 screen.

[SP777]

Originally Posted by dpavlin

Two column mode isn't magic. It assumes that two columns are in the middle of page.

I've already posted on this issue. It's very easy to recognize any number of columns even in scanned documents.

Originally Posted by SP777

There is a good way to recognize any number of columns (not just two) and their exact boundaries. View the page as rasterized bitmap. Sum up the number of active pixels in each pixelwide column of the page. Zero or low values will show all the blank space (page and intercolumn margins).

[hawhill]

SP777: In fact, I did once create a similar algorithm to create a comic book scan slicer. So I have a basic understanding of the principles - although I didn't read up on popular algorithms for that and just started coding :-P (Hacker attitude, lazyness) Hopefully I will get back to porting algorithms like that to our Lua code. We've just switched over to LuaJIT (impressive technology) and doing such heuristics in pure Lua is now a real option performance-wise. Your input is nevertheless very welcome. I agree that there would be real need to optical document structure recognition for providing a good PDF reading experience on these under-dimensioned (for the task) e-readers.

[dracodoc]

There is nothing wrong with your pdf or kpdf. You can check the help for the 2 column mode here
https://github.com/hwhw/kindlepdfvie...tCutsForReader

F will move to the top left to show the left column. Press down to get to bottom of column and then right to get to top of next column to the right). Position will be reset to top-left on each page change.

Originally Posted by lesen54321

Hello,

lesen54321

[dracodoc]

Reflow is not easy, or even impossible for some pdf with complex tables, graphs, codes, etc.
Just crop the edge, use landscape mode will be enough for most of pdf.
If really need to read complex pdf in small screen reader, there are software like k2pdfopt, although I never use it since I use pdflrf with dxg extensively.

On the other side, the "scroll mode" will be very useful and possible to be implemented, which means connect all pages into a continuous scroll, fit page width or content width, every page down will move to next screen, and pages are continuously connected.

[reprep]

i agree with dracodoc as the "scroll mode" would be a great feature for this wonderful viewer.

[hawhill]

Yes, I'm aware of that. There is an issue in the tracker now (thank you dracodoc!) where a bit of discussion can take place. It's a bit harder to implement than it seems at first sight, though. It won't be forgotten, but please don't expect it right away... (Although the developer community around this project has often surprised me with its inspiration and work.)

[bloor]

Originally Posted by bloor

Fantastic stuff so far

However with the new update, large pdfs (~100 mb) page turn/commands are painfully slow (upto 15-20 seconds+), I'm assuming its the limited RAM on the dxg

Shift PK doesnt seem to be enough, is there anything else I can do on my end?

-Thank you

Bumping this (correction: file is 400 mb ) , using the stock pdf reader with no issues atm