[GUI Plugin] Kobo Utilities

[jackie_w]

Originally Posted by ownedbycats

Wondering purpose of option to convert covers to black&white before upload. Was put in when Kobo just BW anyways.

Originally Posted by Quoth

Just to save space. Might speed loading slightly?

The B&W covers option is not just about saving space. It's also a prerequisite for other KU (and KoboTouch driver) cover options:

• Upload dithered covers
• Save covers as PNG

which enable much improved cover image dithering, i.e. reducing the ugly "banding" seen on the sleep screensaver for some books.

[JSWolf]

Originally Posted by ownedbycats

Ah, was wondering if maybe it rendered different.

It may dither better then Kobo did back when the option was created.

[ownedbycats]

Originally Posted by jackie_w

which enable much improved cover image dithering, i.e. reducing the ugly "banding" seen on the sleep screensaver for some books.

Ah, the effect you get when you reduce number of colours in image? I not see it, maybe newer firmware handles better.

[Quoth]

Originally Posted by jackie_w

The B&W covers option is not just about saving space. It's also a prerequisite for other KU (and KoboTouch driver) cover options:

• Upload dithered covers
• Save covers as PNG

which enable much improved cover image dithering, i.e. reducing the ugly "banding" seen on the sleep screensaver for some books.

I've never dithered cover image, so never seen banding.
Used covers "as is" on Kobo Touch, Original Aura H2O, Libra, Libra2, Sage and Elipsa.

Dithered would want to be PNG, as JPEG can make a mess of dithering anyway. A dithered image is a disaster to rescale, so a dithered image only suits something than can do the source to display pixels 1:1.

[Quoth]

Originally Posted by ownedbycats

Ah, the effect you get when you reduce number of colours in image? I not see it, maybe newer firmware handles better.

No, "banding" in display of a dithered image isn't due to reducing number of colours (though badly done that can create posterisation effects). Dithering allows using less display colours (or 1 bit black & white) to show all the source shades. Banding is a Moire effect due to not displaying the dithered image pixels 1:1 with display pixels.

[JSWolf]

As my Kobos are not color, I'll leave this option check to pre-dither the cover images. I've not seen any issues.

[jackie_w]

Originally Posted by Quoth

I've never dithered cover image, so never seen banding.
Used covers "as is" on Kobo Touch, Original Aura H2O, Libra, Libra2, Sage and Elipsa.

Dithered would want to be PNG, as JPEG can make a mess of dithering anyway. A dithered image is a disaster to rescale, so a dithered image only suits something than can do the source to display pixels 1:1.

It's a long time since @NiLuJe/@davidfor explained this but IIRC...

If your calibre covers are colour (almost certainly), even if you 'do nothing', dithering happens somewhere along the line between calibre send-to-device and the Kobo displaying the sleep screen cover because the Kobo device itself will dither images whenever it thinks it needs to. @NiluJe's opinion (he was/is very picky about cover image quality) was that leaving it up to the Kobo device gave the worst end results. Hence he supplied the code for the extra cover options in KU/KoboTouch driver and @davidfor implemented them.

Enabling ALL the cover options (Keep aspect ratio, Letterbox fullscreen covers, Send B&W, Upload dithered covers, Save covers as PNG) ensures that all dithering for the sleep fullscreen image happens at the calibre end prior to sending, rather than leaving it to the tender mercies of the Kobo device. I believe using the PNG format allows dithering to the correct eink 16-colour grayscale palette, whereas JPG does not.

In summary:
If before it's sent, the fullscreen image:

• is already B&W using only the 16-colour eink grayscale palette
• already has the 2 vertical padding sidebars added in your colour of choice
• is already the exact dimensions for the Kobo model it's being sent to

then the Kobo will not need to do any further dithering.

[compurandom]

Dithering is a method for converting high pixel depth low resolution into a format suitable for display on high resolution low bit depth devices and formats.

Typically you would use it to convert (for instance) a 72dpi 8 bit per color image for a screen into a 600dpi 1 bit per color image for a printer. The trade off of resolution for bit depth lets you simulate higher bit depth on the printer and look about the same. Dithering the image to a resolution different than the device resolution doesn't really make a lot of sense, and will cause moire banding and doing it at a resolution different than the device resolution is basically wrong and worse than doing nothing.

My assumption is that the original kobo was only black and white, with no grey scale. You can simulate grey scale with dithering.

Also, I assume that pre-dithering the image for a screen that already supports grey scale is probably obsolete. A screen with 4 bit depth (16 shades of grey) is not bad and dithering to black and white will probably make it look worse, although it would be possible to dither 8 bit to 4 bit rather than 8 bit to 1 bit and have it look ok. But then, I would also assume that the device knows its gamut better than a generic dithering algorithm, and letting the device dither to its version of grey scale is probably better.

I would also assume that the 4 bit grey kobo screen is really just black and white dithered to simulate 4 bit depth, and that dithering twice probably isn't that great either.

[Quoth]

Originally Posted by compurandom

…
Dithering the image to a resolution different than the device resolution doesn't really make a lot of sense, and will cause moire banding and doing it at a resolution different than the device resolution is basically wrong and worse than doing nothing.

My assumption is that the original kobo was only black and white, with no grey scale. You can simulate grey scale with dithering.

Also, I assume that pre-dithering the image for a screen that already supports grey scale is probably obsolete. A screen with 4 bit depth (16 shades of grey) is not bad and dithering to black and white will probably make it look worse, although it would be possible to dither 8 bit to 4 bit rather than 8 bit to 1 bit and have it look ok.
…

I agree, I can't see the point of pre-dithering for Kobo as it's black, white and 14 greys and you have to get the resolution / scale exactly right.


There were early Sony & Kindle eink with only 2 bit levels. I'm sure that was before Kobo did any eink and all mono Kobo are 4 bit levels. All Pearl screens and all later Carta are all 4 bit. Later Vizplex (the one before Pearl) are 4 bit.

Hercules graphics or CGA mono was only 1 bit.
Most of the Sony Data Diskman were 1 bit LCD.

[JSWolf]

Originally Posted by Quoth

I agree, I can't see the point of pre-dithering for Kobo as it's black, white and 14 greys and you have to get the resolution / scale exactly right.


There were early Sony & Kindle eink with only 2 bit levels. I'm sure that was before Kobo did any eink and all mono Kobo are 4 bit levels. All Pearl screens and all later Carta are all 4 bit. Later Vizplex (the one before Pearl) are 4 bit.

Hercules graphics or CGA mono was only 1 bit.
Most of the Sony Data Diskman were 1 bit LCD.

What I know is the dithering for the cover images works well. It works well on my H2O and works well on my Libra 2. Different resolutions but no issues with the covers.

[DNSB]

For what it may be worth, the last time I saw any difference between using KU to dither the images and Kobo's builtin dithering was before firmware revision 4.13.12638 back in 2019. I'd been using KU to send cover images for years by then but did some testing after noticing I could not see a difference in the images in my ePub tester ebook in that version. It may well have happened earlier, this was just the first time I noticed the lack of difference. This may also say something about how rarely I see full page colour images in a book that I am reading.