Solution: Monochrome PDFs which shows missing charecters

[CapitainDerya]

For Kindle Paperwhite 2 and Basic;
I noticed that when i view a pdf which is created using monochrome-scanned images , It doesn't show some characters. Solution is to convert pdf to grayscale format.

But Is it possible to tweak the kindle pdf reader in order to fix this issue?
I think that it should display monochrome images properly, because it is smaller and optimized.

[Cinisajoy]

Are you using a factory kindle or did you jailbreak it?

[eschwartz]

No idea, I don't really read PDFs much.

But you might want to give KOReader a try, if you are jailbroken.
It is a much better PDF reader on general principle.

[geekmaster]

Some "fast update" eink modes display all grayscale values as white, usehull for "animation". like a mouse cursor for example. Also when stretching or zooming text, then fill in the invisible grayscale by redrawing the last static image with a slow eink update mode.

If your grays are missing (and some antialiased characters migh not contain any pure black pixels), you may be using a no-flash (fast animation) eink update mode.

Then again, it could be a software bug. What app are you using to view these PDF file images?

[CapitainDerya]

I uploaded and updated images above in the first post. I also uploaded some monochrome pdf. If you convert the images embedded to pdf to grayscale , it will not display missing characters

[geekmaster]

Originally Posted by CapitainDerya

I uploaded and updated images above in the first post. I also uploaded some monochrome pdf. If you convert the images embedded to pdf to grayscale , it will not display missing characters

I just did a screen capture and histogram analysis of that PDF page, and it contains 32 evenly spaced gray value spikes, with low-level scattered pixels of other grayscale values. It is possible that some letters contain an "almost black" but no pure black. I will look deeper...

[CapitainDerya]

Originally Posted by geekmaster

I just did a screen capture and histogram analysis of that PDF page, and it contains 32 evenly spaced gray value spikes, with low-level scattered pixels of other grayscale values. It is possible that some letters contain an "almost black" but no pure black. I will look deeper...

You mean, missing characters are not pure 0x00(Black). I will edit via photoshop in order to convert these to the pure black(0x00) and check if it will display correctly

[geekmaster]

Now this is weird. After viewing the histogram (attached image) I convereted it to 8-bit grayscale so I could view and edit the color palette (also attached below). Although the histogram cleary shows an even distribution of pixels that are predominantly 32 evenly spaced shades of gray, the color palette seems to show only 16 colors, and viewing the histogram again shows no change -- 32 spike like before.

So, is it 16 colors (plus some digitized noise), or is it 32 colors? Confusing. And those missing letters in your image look the same (visually) as all the other letters. You would thik atleast a few pixels would show. And the PDF appears to be backed by no OCR text, so only an image.

Not only do your missing letters make no sense, but I am also confused by the contention between histogram and color palette. Weird.

[CapitainDerya]

geekmaster, did you extracted image as bmp or png from pdf? I think that this pdf is embedded with monochrome(1 bit) image

[geekmaster]

I reduced the color palette to 32 colors, and the historgram shows the same 32 spikes, but no background noise (flat baseline). Then I reduced it to 16 colors and it had 16 spikes as expected, so the histogram appears to be working and that color palette from the screen capture is just weird. Worse, even though there are only 32 used colors, the new palete after reducing to 32 has 256 shades of gray, not those 16 different colors in that 256 palette attached above.

[CapitainDerya]

I uploaded extracted image via adobe dc

[geekmaster]

Originally Posted by CapitainDerya

geekmaster, did you extracted image as bmp or png from pdf? I think that this pdf is embedded with monochrome(1 bit) image

I did GIF to upload, but I extracted 32-bit pixels with a screen capture. Perhaps that was distorted by the PDF viewer which anitialiased while resising to fit my screen. The correct approach would be to digitally rip the image from the PDF file directly, not a screen capture, but this computer has no such tools.

In any case, a 1-bit image is all pure black and pure white -- no gray pixels. It makes no sense that the kindle would arbitrarily decide to erase some blocks of pixels that just just happen to be a complete letter without affecting adjacent letters. It would need contour detection to selectively "lasso" and erase letters like that.

It is a puzzle I do not wish to solve right now -- I had a very rough day...

[CapitainDerya]

I uploaded in the first post, grayscale version of pdf which displays properly

[geekmaster]

It is strange that the kindle erases some letters in an image embedded in a PDF when not grayscale. That would take smart code to do that, which makes no sense. It is just one big monochrome image. Glad the grayscale works, but that other problem is a puzzle only lab126 or amazon can solve (unless they published the source code that contains this conundrum).

[NiLuJe]

Crazy scaling artifact, perhaps?