1.7.0.107 - Big regress in Quality of Denoise-Filter!

[Polygonius]

Maybe the new Denoise-algorithm is better for pure denoise, but the old denoise-filter was also a subtle kind of lovely blur. I miss the old algorithm for "smooth blurring"

 

Please compare both picture. The new one is really harsh, doesnt create this lovely smoothing. No other filter creates this kind of "bluring". So please give back the old denoiser (maybe with a new name and maybe in the blur-section and maybe with EXTREM up to 10K - so i do not need many instances). 

First is the new one (it muddies all, but not the hair), second the old one (the hair gets nicely blurred). (both  2 times with same settings as shown).

[Puck]

1 hour ago, Polygonius said:

Maybe the new Denoise-algorithm is better for pure denoise, but the old denoise-filter was also a subtle kind of lovely blur. I miss the old algorithm for "smooth blurring"

 

Please compare both picture. The new one is really harsh, doesnt create this lovely smoothing. No other filter creates this kind of "bluring". So please give back the old denoiser (maybe with a new name and maybe in the blur-section and maybe with EXTREM up to 10K - so i do not need many instances). 

First is the new one (it muddies all, but not the hair), second the old one (the hair gets nicely blurred). (both  2 times with same settings as shown).

I'm with you! I'd also like to have both versions of this filter available. I also like to use the previous filter with this smooth way of blurring. No other filter achieves this.

Regards, Puck

[Jowday]

I made a few comparisons without using extreme settings. The old denoise algorithm was not great, but I found it usable. The new denoise algorithm is blurring lines with details or nearby areas of solid surfaces. I have never seen such a bad implementation before. "Makeup with tear drops" Serif could name the brand new denoise feature from Serif Labs. Perhaps I did push it too far though. Never mind.

If you are serious about noise reduction the built-in features inside Photo and Photoshop are not an option. Adobe Camera RAW is kind of ok.

I have used plugins from Topaz and Noiseware Professional (my favorite) for many years, and if the Nik Collection from Google still works with Photo, there is a free competent filter included as well.

A great noise reduction plugin is worth every penny; you can work with higher ISO settings and get great results anyway. Try DxO Photolab fx with Prime Denoise enabled. It is MAGIC.

[Polygonius]

3 hours ago, Jowday said:

If you are serious about noise reduction the built-in features inside Photo and Photoshop are not an option. Adobe Camera RAW is kind of ok.

I have used plugins from Topaz and Noiseware Professional (my favorite) for many years, and if the Nik Collection from Google still works with Photo, there is a free competent filter included as well.

Nope, high quality noise reduction is not my point. There are, as you recommend, very good (sometimes free) alternatives. And i use them too... normally i do my developing/prepare outside AP!

For me AP is a creative-edit-tool, not a darkroom/enhancer.... there are other apps several times better...  

But maybe lot of people will do ALL inside AP?  And why not? ist never a question of either/no! Its just a question of have BOTH possibilities! So the new denoise is maybe better for just denoise... but it lost its "creative usabaility" - i prefefer if my "images" are ready for "edit". 

Again: Is not a question of either/or... its just a simple decision: Give us both! (The algorithm are STILL developed/avaiable, so juts give one a new name/new menue-entry and its gone!).  What do have to lost? 

(I create Audio-FX... often are more/less similar technically developing will create really different effects! Sometimes just the change of a CONSTANT to an un-normally "outrange" will give the FX a complete new flavour!  In this case i add a new button/slider/knob... or split the FX (pseudo-plural) into  separate units, with fresh name (for dedicated developing editing)...  There is NEVER NOTHING to lost... its (if both candidates really differ) always a win! There is no LAW that an effect has to be exactly so or so! Sometimes "mistakes" inspire very much! So whats the question? Both algorithm are already fine (for different jobs - so there is no either/or... there is just a who should i place BOTH? and how to name???... )

Keep one as its "destination-deveolping" and the other one as a "random-gift", but take BOTH... just think about a new name/or how to present both (several "flavours")  "inside" one FX, ähm sorry, "Filter"

 

ABSOLUTELY SAME for the discussion about the harsh displacement-filter. There is a very good reason to overwork this "harsh" quality as a "NEW" filter, but there is also a very good reason to keep the old-one as it is... its a very unique SFX, so make a new "high-quality" one, BUT keep the dirty old one!  

 

For my experience of developing Audio-FX- ... in 60% its not an either/or... its mostly a take both (refine/specialize both), but  how to "present" both better "inside"  a single FX (as "flavor")or create an autark FX )if really "other)...

However, second-hand / random-developing... is a very thankfully pool for new ideas...  A "drawing board" just creates (in best case) what you imagine... A "random" does not create what you imagine, but sometimes a very good "new item"... like here: The old denoiser will not be the best denosier... but you have created randomly a very nice, unique "blur". Be proud and give us the option to use this "imperfect" random one (however you call/categorize) AND the "real" one, you have "drawing-bord-developed".

But there is absolutely no reason to trash a good-random FX... Nope, its on the user to use both/several or just one... 

Whenever there is a NICE random-"artifact", keep this, as ready or as "potential", develop further.. item.

 AGAIN: You can not lost (except some HD-space), just win! 

 

[Jowday]

I get your point ... but then it really should be designed as a creative filter and labeled as such. With the ability to label and save your preferred setting.

The majority if users will expect a denoise filter designed to reduce ISO noise patterns or JPG artifacts most of all. The denoise filter must excel in that area.

[Puck]

6 hours ago, Polygonius said:

The old denoiser will not be the best denosier... but you have created randomly a very nice, unique "blur". Be proud and give us the option to use this "imperfect" random one (however you call/categorize) AND the "real" one, you have "drawing-bord-developed".

But there is absolutely no reason to trash a good-random FX... Nope, its on the user to use both/several or just one... 

Whenever there is a NICE random-"artifact", keep this, as ready or as "potential", develop further.. item.

 AGAIN: You can not lost (except some HD-space), just win! 

I agree 100%.

Regards, Puck

[Chris B]

Thanks for your feedback everyone.

I think keeping both would be a bigger job than you may think. Not only would we require programming changes but we would also need documentation, tutorials, Help topics and translation. That's a lot of work for something that has essentially been fixed (but granted is different).

I'll pass on what has been discussed.

 

[fde101]

3 hours ago, Chris B said:

I think keeping both would be a bigger job than you may think. Not only would we require programming changes but we would also need documentation, tutorials, Help topics and translation. That's a lot of work for something that has essentially been fixed (but granted is different).

You could minimize that by making the "old" algorithm activate using a checkbox on the existing window ("Creative effect" or whatever) instead of trying to treat it like an entirely separate filter.  This would also facilitate importing older files by enabling the checkbox by default when the older files are imported, as that would prevent the appearance of older documents from being changed when someone imports them into a newer version, something that really shouldn't happen.

[Puck]

14 hours ago, fde101 said:

You could minimize that by making the "old" algorithm activate using a checkbox on the existing window ("Creative effect" or whatever) instead of trying to treat it like an entirely separate filter.  This would also facilitate importing older files by enabling the checkbox by default when the older files are imported, as that would prevent the appearance of older documents from being changed when someone imports them into a newer version, something that really shouldn't happen.

Very good objection and reply! 

Regards, Puck

[Jowday]

20 hours ago, Chris B said:

Thanks for your feedback everyone.

I think keeping both would be a bigger job than you may think. Not only would we require programming changes but we would also need documentation, tutorials, Help topics and translation. That's a lot of work for something that has essentially been fixed (but granted is different).

I'll pass on what has been discussed.

 

I wouldn't call it fixed. Replaced. Different. I hope it was replaced because of a critical mass of unhappy user feedback. You are also changing workflows and results out there in the world. Customers are confused or blocked from getting the results they are used to. Now THAT is a lot of work that is influenced. Outside Nottingham.

It is a dangerous approach to software development to replace on technology/algorithm with another one without notice. You have done it before. It is not optimizing. It is replacing. Retreat. One could also argue that with so many years of experience (Photoplus and now APhoto) you should have had enough in-house knowledge to have chosen a proper, mature denoise algorithm years ago in Photo.

Noise reduction is tricky to get right and it takes hundreds of image processing hours to really nail an intuitive understanding of the best settings for an image or layer. The same goes for sharpening etc. All this knowledge your users acquire can be lost in a minor update.

[fde101]

24 minutes ago, Jowday said:

It is a dangerous approach to software development to replace on technology/algorithm with another one without notice. You have done it before. It is not optimizing. It is replacing.

In many cases this is a good thing.  Improving a product is better than not doing so.

However, in this case it is a replacement that is producing different results for existing documents, meaning that work that has already been done will be rendered differently in a newer version of the product.

That is a problem which should not be taken lightly.

 

EDIT: to add to this, when Apple was producing Aperture, they made a big deal that if they improved their RAW processing engine for a photo, they would include all of the old versions of the engine along with the new version and allow the user to switch among them.  This would provide for older photos being rendered in exactly the same way that they were when the user first processed them.  If the user wanted to leverage the new engine they would simply pick it in a list when revising the photo, but any photos they had already processed would not be altered without their review.  Any truly "professional" application needs to take things like this seriously.

[Chris B]

If you open an afphoto file with a Live Denoise Filter, it will use the old method (old UI as well) so this shouldn't really be a problem... The change only applies to a new Live Denoise Filter in which I am able to get the same results as the old method albeit with different positions on the slider. 

You will have to slightly retain yourself to take into account the different in slider positions etc.

Here's a screenshot with a 1.6 afphoto file opened in 1.7. Notice how it has the same Live Denoise dialog as in 1.6 and the filter appears unchanged.

[fde101]

Hmm...  if the argument for not keeping the old one around is the need to maintain extra documentation, and the older files open using the same interface as the older version of the product, then the functionality of the older filter is still there anyway, and the documentation would still be needed to avoid confusing people who receive files from others that were created using the prior versions?

It also means that the only thing missing to provide both would be the interface to create it.

If it is truly the case that all variants of the old behavior can be reproduced using the new filter, wouldn't it be better to simply translate the slider positions to compatible placements on the new version during import so that you have only one version to document and maintain?

[Puck]

15 hours ago, fde101 said:

Hmm...  if the argument for not keeping the old one around is the need to maintain extra documentation, and the older files open using the same interface as the older version of the product, then the functionality of the older filter is still there anyway, and the documentation would still be needed to avoid confusing people who receive files from others that were created using the prior versions?

It also means that the only thing missing to provide both would be the interface to create it.

If it is truly the case that all variants of the old behavior can be reproduced using the new filter, wouldn't it be better to simply translate the slider positions to compatible placements on the new version during import so that you have only one version to document and maintain?

That are good arguments. I think it makes a lot of sense to keep the old filter (the algorithm is obviously still present in 1.7). Good suggestions have already been made here, the developers should take them to heart.

[Polygonius]

On 1/28/2019 at 3:26 PM, Chris B said:

.... but we would also need documentation, tutorials, Help topics and translation....

Is this really an argument against a "new" (in this case  "old") feature? You have to rewrite/translate the new manual so or so, that s just one small chapter more. And its not true: you neither have tutorials/videos to all aspects, nor it is necessary to have this stuff from the company. There are lot of discover-agents outside, which explain "hidden" features very well in the www... Power-users just need a rudimentary description - the rest they will discover by them-self.

"This new BLUR works like the old version of denoise... It works like the opposite to the bilateral-blur: It blurs edges/lines and keep the rest more/less intakt, during the bilateral-blur works in quite the opposite way."

 

BTW: If you had made a PDF-help instead this incomfortable system-help, it would be maybe much easier to rewrite the manual. You had just to write and export, not to "programming" it! And for the user a chapter-collapse PDF is much more comfortable than this kind of help. 

 

BTW 2: Because translation - for me it would be nice if there ALWAYS where a mixmatch between GENERAL native and common international terms: For example i would like that tools, adjustments, filter, blendmodes... and other very common INTERNATIONAL terms keep the COMMON english names, just the menue and the help should be translated. And i thing i´m not alone. Often the english terms are in several languages much more used than the native-ones in specific situations. Noboday say "Verzögerung" in music-production, all says "delay". Nobody says "Versatz" in photo-apps, all says displacement....