PDFRead 1.8.2 released!

[nrapallo]

Originally Posted by =X=

@ashkulz Here is a link This file is 17MB because it is already and image.

Also this effect is not tied to a specific PDF, it is observed on any given PDF,
=X=

OK, don't mind me butting in ( it IS my thread ), but I've made a simple change to PDFRead's use of the MinFilter to acheive better results on that .pdf file provided, IMHO.

The MinFilter can use a 3x3 matrix (default=3) or 5x5 matrix when doing its magic. I changed the dilation module's return value in process.py to:

  return image.filter(ImageFilter.MinFilter(5))

That now results in thicker text prior to the resize which uses the Lanczos filter. However, on low-res .pdf it causes the text to blur too much. In that case, raise the initial 300 DPI to 400 or 500; 600 was very slow, but may be better?

Could you test to see which resulting .lrf (first 5 pages only) looks better on the hardware reader? (Sorry for the stretch of short pages, I still have to fix that one!!!) What about the comparison with PDFLRF now?

p.s. I provide the modified PDFRead bin files which should be placed in the bin directory and overwrite the existing files. Just take a backup first though should you wish to restore those original files.

[=X=]

Originally Posted by nrapallo

OK, don't mind me butting in ( it IS my thread ),

Well considering it was a productive meddling you're excused

Okay I've taken a look at the files, here's the good, the bad.

There is a definite improvement on these files over what 1.8 produces today. They show up much clearer on the reader. There where two files that looked better than the rest. On was the first file minifilter-3 and the other was minifilter-5-400


What I see in PDFReader is a truncation of fonts on the top or bottom part. The best example is on Pg 6 of the landscape. However this occurs through out the text.
Some of the pixels on the font are getting erased, by dropping the dilation you overcome this but it bleeds over. I think if you stretch the page out the bleeding will be reduced

PDFLRF still looks better but again it is maximizing the screen by croping off the white spaces. I do notice PDFLRF does bleed seen mostly with the "a" font. I suspect if you dilate with a value of 300 and stretch the screen you should get similar results to that of PDFLRF.

=X=

[nrapallo]

Originally Posted by =X=

Well considering it was a productive meddling you're excused

Okay I've taken a look at the files, here's the good, the bad.

There is a definite improvement on these files over what 1.8 produces today. They show up much clearer on the reader. There where two files that looked better than the rest. On was the first file minifilter-3 and the other was minifilter-5-400

Is it Coke or Pepsi? Years ago, this was the norm for commercials by Pepsi where the "public" was asked which one was better (without seeing the brand). On the Pepsi commercials, all chose "Pepsi"!

Ok back to reality, the MinFilter-3 is the original PDFRead executable with the default MinFilter(3) with only 8 colors. The MinFilter-5-400 is the MinFilter using the 5x5 matrix and 400 DPI. Here you get a higher resolution with thicker text before the resize op kicks in.

What I see in PDFReader is a truncation of fonts on the top or bottom part. The best example is on Pg 6 of the landscape. However this occurs through out the text.
Some of the pixels on the font are getting erased, by dropping the dilation you overcome this but it bleeds over. I think if you stretch the page out the bleeding will be reduced

That's just the anti-aliasing done in 8 colors; try using 4 colors or 2 colors! The edges are smoothed between the black text and white background and anytime the transition is more white/gray you get that "loss" effect.

PDFLRF still looks better but again it is maximizing the screen by croping off the white spaces. I do notice PDFLRF does bleed seen mostly with the "a" font. I suspect if you dilate with a value of 300 and stretch the screen you should get similar results to that of PDFLRF.

=X=

The pylrf library used by PDFRead handles all this for the program, so that stretching adds further to the distortion of the text and more lost pixels. Now to figure out a way to counter-act it...

p.s. However, PDFRead does produce a page which is cropped to remove the margin, so maybe the pylrf library is adding that white margin? Have you seen the .png pages that PDFRead produces? If not, use that 'debug' feature explained in post #1 above (last entry under [2008-03-12] 1.8 (by AK and NR)).

[ashkulz]

Looks like the Sony software has issues with VirtualBox, so I'll have to download and install VMWare

[=X=]

Originally Posted by nrapallo

Is it Coke or Pepsi?

Okay that is what I get for trying to talk intelligently beofre my

Originally Posted by nrapallo

The pylrf library used by PDFRead handles all this for the program, so that stretching adds further to the distortion of the text and more lost pixels. Now to figure out a way to counter-act it...

p.s. However, PDFRead does produce a page which is cropped to remove the margin, so maybe the pylrf library is adding that white margin?<snip...>

I think your on to something. The PNG images produce from PDFRead do look much better than the final LRF file.

I've attached an LRF created by Calibe/BookCreator (I simply slapped the images in a word document then stretched them out.

The result was pretty decent.

=X=

[nrapallo]

Originally Posted by =X=

Okay that is what I get for trying to talk intelligently beofre my

Sorry, I couldn't resist!

But, I do I appreciate the amount of effort your have expended here with this issue.

I think your on to something. The PNG images produce from PDFRead do look much better than the final LRF file.

I've attached an LRF created by Calibe/BookCreator (I simply slapped the images in a word document then stretched them out.

The result was pretty decent.

=X=

I had previously looked at Kovid's modified pylrf modules, but couldn't figure out what to change to 'mimic' his lrf generation. Python coding is NOT my forte, so I just tried superficial changes and they all failed. Oh well, I do know someone who knows python very well...

I liked page 1 of your latest .lrf posting (viewed with the software reader). Page 4 was nice too but appeared a bit too squished and lost some definition. The last page was a write-off. Which one did I favour. Was it Coke?

[=X=]

Originally Posted by nrapallo

Was it Coke?

Nope it was Pepsi

=X=

[nrapallo]

Originally Posted by =X=

Nope it was Pepsi

=X=

Pepsi (MinFilter-5-400)?

[=X=]

Originally Posted by nrapallo

Pepsi (MinFilter-5-400)?

PDFRead Dilation 300 DPI (All the images where produced from PDFRead)

[nrapallo]

Originally Posted by =X=

PDFRead Dilation 300 DPI (All the images where produced from PDFRead)

From the modified PDFRead bin (for that MinFilter-5 recent change); all at 300 DPI right?

What would 400/500 DPI look like then.

[=X=]

Option e) "None of the above"

It's an unmodified PDFRead v1.82. I did change the settings from my very first post, as I've been learning how to use PDFRead

Settings;
DPI: 400
color: 4
Crop %: 5
Edge Level: 6
H/W: 754/584
The rest of the settings are default

Now I'm tempted to see how good the modified PDFRead looks.
=X=

[nrapallo]

Originally Posted by =X=

Option e) "None of the above"

It's an unmodified PDFRead v1.82. I did change the settings from my very first post, as I've been learning how to use PDFRead

Settings;
DPI: 400
color: 4
Crop %: 5
Edge Level: 6
H/W: 754/584
The rest of the settings are default

The Edge level is quite high. 5 is better or try checking the no edge enhance box at the box of the GUI. BTW, the default prs505-p profile uses 753/583 as it's H/W.

Now I'm tempted to see how good the modified PDFRead looks.
=X=

I hope you are pleasantly surprised when you do. Quick, buy some PDFRead stock, now.....

[=X=]

So yesterday I tried this method to make PDF readable under the SONY. The process is found under Poor boys way of editing PDF.... Anyhow with SONY's firmware release for the PRS-505 this solution is very nice. The advantage is that it keeps the PDF as text and not into an image.

Essentially this tool does two things, remove the margin and change the fonts. I found removing the margin the only thing necessary, but changing the font also helped . I guess the reason I'm posting here is to wonder how hard it would be to include this as an additional SOLUTION for PDF?

=X=

[nrapallo]

Originally Posted by =X=

...an additional SOLUTION for PDF?

=X=

Other possible solutions as posted by caritas (uses Sony PRS505) include:

An algorithm to render PDF in small devices

Subtitute font when rending PDF on small screen

Another embold method for rending PDF

They all have stengths and weaknesses, but since PDFRead is an image based solution right now, I would think the one in bold, above, would be the BEST solution overall!

[=X=]

Well yes they are all good. (The middle solution is modifying the font on the PDF before it is converted to an image hm...)

The problem with image based solutions is that they loose font crispness as so do the images in the PDF. This is do to all the image manipulation done get the image viewable on the target device. Also the final eBook is HUGE compared to the original. A typical 500KB PDF will be close to 10MB.

Now only modifying the margin and Font will preserve the images and font quality in addition to keeping the final eBook size very small

Anyhow just food for thought

=X=