I just got a new computer from Newegg. It makes my old one look like junk. With 8 cores instead of 2, a brute force optimization takes 1 day instead of 4.

I have noticed that even though a cubic predictor has a higher PSNR, a quadratic predictor looks better for lossy compression because it smudges edges less. Smudges are more annoying than other artifacts. Even better results can be achieved by detecting which blocks have an edge and using more bits.

For this image:
https://www.pngegg.com/en/png-wvrud

here are the file sizes:

asymmetric cubic 4099820
symmetric cubic 4035623
symmetric quadratic 3966858

I use coffee to give me energy to program. I want internetters to watch me develop the best image compressor.

--
Get stewed.
Books are a load of crap.

On Sun, 22 Jan 2023 22:09:06 -0800 (PST)
Mihai Moise <mihai.moise5@gmail.com> wrote:

> I just got a new computer from Newegg. It makes my old one look like junk. With 8 cores instead of 2, a brute force optimization takes 1 day instead of 4.
>
> I have noticed that even though a cubic predictor has a higher PSNR, a quadratic predictor looks better for lossy compression because it smudges edges less. Smudges are more annoying than other artifacts. Even better results can be achieved by detecting which blocks have an edge and using more bits.
>
> For this image:
> https://www.pngegg.com/en/png-wvrud
>
> here are the file sizes:
>
> asymmetric cubic 4099820
> symmetric cubic 4035623
> symmetric quadratic 3966858
>
> I use coffee to give me energy to program. I want internetters to watch me develop the best image compressor.
>
The coffee makes you post the same thing over & over?

--
Bah, and indeed Humbug.

"Kerr-Mudd, John" <admin@127.0.0.1> writes:
> On Sun, 22 Jan 2023 22:09:06 -0800 (PST)
> Mihai Moise <mihai.moise5@gmail.com> wrote:
>
>> I just got a new computer from Newegg. It makes my old one look like
>> junk. With 8 cores instead of 2, a brute force optimization takes 1
>> day instead of 4.
>>
>> I have noticed that even though a cubic predictor has a higher PSNR,
>> a quadratic predictor looks better for lossy compression because it
>> smudges edges less. Smudges are more annoying than other
>> artifacts. Even better results can be achieved by detecting which
>> blocks have an edge and using more bits.
>>
>> For this image:
>> https://www.pngegg.com/en/png-wvrud
>>
>> here are the file sizes:
>>
>> asymmetric cubic 4099820
>> symmetric cubic 4035623
>> symmetric quadratic 3966858
>>
>> I use coffee to give me energy to program. I want internetters to
>> watch me develop the best image compressor.
>
> The coffee makes you post the same thing over & over?

His redundancy makes him easier to compress.

Phil
--
We are no longer hunters and nomads. No longer awed and frightened, as we have
gained some understanding of the world in which we live. As such, we can cast
aside childish remnants from the dawn of our civilization.
-- NotSanguine on SoylentNews, after Eugen Weber in /The Western Tradition/

Is midnight 0:00 or 24:00? I never understood it. Maybe other people can explain it to me.

On 1/30/23 9:33 PM, Mihai Cartoaje wrote:
> Is midnight 0:00 or 24:00? I never understood it. Maybe other people can explain it to me.

Conventional rules say the hour is 00 - 23, so 00:00

You might be able to argue that the exact instant of midnight is 24:00
since that is the end of the day, but the very next moment would be

00:00:00.0000...01 since it IS in the next day, so 0 hours into the day.

So we don't write Midnight as 24:00 as that is only plausable for an
infintesimal fraction of a second.

Libima is the alternative to WiMP.

I can post about non-compression topics in this thread. It's mine.

I can post about non-compression topics in this thread. That's mine.

You people don't appreciate my posts.

I realized why they are called bookshelf speakers. The books absorb the echo.

I brew rosehip tea and coffee together. I like coffee cold, like the iced cappuccinos at Tim Hortons. With breakfast eggs. Since Google adopted a javascript interface, this newsgroup is dead. So no one is watch me write how I program.

I brew rosehip tea and coffee together. I like coffee cold, like the iced cappuccinos at Tim Hortons. With breakfast eggs. Since Google adopted a javascript interface, this newsgroup is dead. So no one is watch me write how I program.

I am also good at physics. Though not a crackpot like the people who want the goverments to fund less experimental physics and more people who don't believe in quantum mechanics. When I read a physics book, I notice the errors experts made. I wish I met a hot girl accross the hall.

I don't like archivers which don't have a keep broken option. If a file fails the checksum test, there might still be a reason to keep it. If the file is a video file, the user might want to view the beginning.

I hope you appreciate all the information in my first post. It took me a lot of effort.

The people in London, Ontario say very interesting things just standing in line. Such as,

"Don't use your girl power."
"I'm not short; I'm vertically challenged."

I wonder what percentage of girls in statistics would have a boyfriend.

Or want their parents to like them.

I have invented a lossless 8-DCT which uses 8 multiplications. The previous record was 15 multiplications.

The S transform is

b -= a
a += b >> 1

I shall call the T transform

a += b
b -= a >> 1

A normalized rotation can be made like,

x += c0 * y
y += c1 * x
x += c2 * y

The DCT does not have to be normalized because the scaling can be absorbed in the quantization. If a rotation does not have to be normalized, it frees up a degree of freedom, and the rotation can be made with 2 multiplications.

The last step is inserting these techniques in this flowchart:
https://fgiesen.files.wordpress.com/2013/11/plonka_dct1.png
--
"I can wear whatever you wish. Be anything you wish. [...] You can have whatever dream you want. [...] Let me... please you."

Richard Damon <Richard@Damon-Family.org> writes:
> On 1/30/23 9:33 PM, Mihai Cartoaje wrote:
>> Is midnight 0:00 or 24:00? I never understood it. Maybe other people
>> can explain it to me.
>
> Conventional rules say the hour is 00 - 23, so 00:00
>
> You might be able to argue that the exact instant of midnight is 24:00
> since that is the end of the day, but the very next moment would be
>
> 00:00:00.0000...01 since it IS in the next day, so 0 hours into the day.
>
> So we don't write Midnight as 24:00 as that is only plausable for an
> infintesimal fraction of a second.

ISO thinks otherwise:

"Thus the value ‘24:00:00’ is chosen to represent the final
instant of the day"

"23:59:58’, ‘23:59:59’, ‘23:59:60’, ‘24:00:00’ are times before or at
the ending of the day;"

You need a sensible reason to use it, and it might not be compatible
with tighter standards that utilise ISO8601 (such as some internet
standards), but sometimes it's exactly what you want to use.

Phil

--
We are no longer hunters and nomads. No longer awed and frightened, as we have
gained some understanding of the world in which we live. As such, we can cast
aside childish remnants from the dawn of our civilization.
-- NotSanguine on SoylentNews, after Eugen Weber in /The Western Tradition/

There are people who want Lena to no longer be used. It's nice that they pursue their dream no matter what.

I prefer 0:00, 1:00 , 2:00 than 24:00, 1:00.

I prefer 0:00, 1:00, 2:00 than 24:00, 1:00.

I am not afraid of what people shall think of me from my newsgroup posts.

I'm in Heaven
With my girlfriend

Range Asymmetric Numeral System has the efficiency of range coding and the speed of Huffman.

--
"No. I'd never do anything to hurt you. I want you to come back to me."