Phozn - Fine Tune Your Colors

[danilop]

Hi!

I developed this photographic tuning application and I'd like to share it with you.

The basic idea of Phozn is to divide an image in different zones, usually called segments, based on Hue, Saturation or Value from the HSV Color Model and then apply specific tunings to each segment.

Just load a photo, choose the analysis engine and how many segments you want your photo to be divided into and then apply your specific tuning to each segment:

• The Cluster Analysis engine uses complex algorithms based on the Cluster Analysis theory to divide the image in segments using the HSV Color Model. The relative weight of how Hue, Saturation and Value are taken into account in creating the segments can be specified in the Analysis Options panel.
  
• The Equal Intervals engine simply divides the image in equal intervals of Hue, Saturation or Value as specified in the Analysis Options Panel. Using Equal Intervals with Interval Type Value and 11 Sets is similar to Ansel Adams' Zone System editing approach.

For more informations:
http://www.phozn.com/

There is a basic tutorial with some samples here:
http://www.phozn.com/pages/tutorial

Any positive/negative feedback is welcome!

Rgrds,
Danilo.

[Mark D Segal]

Hi!

I developed this photographic tuning application and I'd like to share it with you.

The basic idea of Phozn is to divide an image in different zones, usually called segments, based on Hue, Saturation or Value from the HSV Color Model and then apply specific tunings to each segment.

Just load a photo, choose the analysis engine and how many segments you want your photo to be divided into and then apply your specific tuning to each segment:

• The Cluster Analysis engine uses complex algorithms based on the Cluster Analysis theory to divide the image in segments using the HSV Color Model. The relative weight of how Hue, Saturation and Value are taken into account in creating the segments can be specified in the Analysis Options panel.
  
• The Equal Intervals engine simply divides the image in equal intervals of Hue, Saturation or Value as specified in the Analysis Options Panel. Using Equal Intervals with Interval Type Value and 11 Sets is similar to Ansel Adams' Zone System editing approach.

For more informations:
http://www.phozn.com/

There is a basic tutorial with some samples here:
http://www.phozn.com/pages/tutorial

Any positive/negative feedback is welcome!

Rgrds,
Danilo.

Danilo,

I looked at your download and testing conditions. I won't do it. I'm not willing to trial-test software unless I have full functionality of all the tools and the final results in 16 bit TIFF for the longer of at least two weeks or until you release a commercial version which is no longer at the beta stage. Secondly, without assurance from experience that it won't do any collateral damage to any aspect of my operating system, there is risk which is not worth taking. Thirdly, I'm not clear how this works in terms of relationship with Lightroom or Photoshop - is it completely stand alone. Does it work with layers, how will it be recognized if the image is opened in other applications after being treated in yours etc.

[danilop]

Mark,

thanks for your comments. This is why I proposed my application to a community of experts as this forum is.

I'm not a legal expert so I used a quite standard license agreement. Do you have an idea of where I can find some conditions that would better fit your expectations?

As I wrote in the download section, if anyone is interested in testing the application without restrictions he has just to get in touch. The only restriction is in the size (not the format) of the saved images, tough.  I'm planning to end this Beta phase at the end of March 2011. I still don't have a commercial plan after that.

I can assure there is no harm to your operating system in using Phozn. It is developed in Java and among the benefits there is the fact that it runs in a sort of "sandbox" (i.e. the Java Virtual Machine).

Phozn can get in input 8/16 bit TIFF or 8 bit JPEG images. It can produce in output the very same formats, so you can have a completely 16 bit Digital Workflow using the TIFF file format.

It is a standalone application and one of the best use case is for a final tuning after a RAW converter. But it can be used after a more complex Photoshop-based editing as well. An output image can be used as a layer in Photoshop and mixed with the original image to create more complex effects. I'm planning a tighter integration with Photoshop and Lightroom in the future but I'm in an early stage right now.

What is different in respect to other image editors is the ability to automatically segment an image in different sets (or zones, if you prefer) in respect to Hue, Saturation or Value of the pixels (as per the HSV Color Model). E.g. in the tutorial I pointed in my previous message you can "select" the sky from the rest of the picture just asking to run the "cluster analysis" engine on hue with 2 sets (one will be the sky, one will be everything else).

With 3 sets you could easily segment an image between the blues (sky, water), the greens and the browns. And so on. There is currently no concept of "locality" in the segmentation algorithms right now, but I'm thinking about that.

I know Phozn is not very user friendly right know, this is why I'm asking for some feedbacks (on the functionalities, on the user interface, etc).

Thanks again for your contributions!

Rgrds,
Danilo.

[feppe]

Love the concept! I'm a huge fan of Lightzone's zone system, which is much more intuitive and easier to use than curves.

Tested it out quickly, here comments.

• Java is not known for speed, but this is VERY slow, taking ~45 minutes to do cluster analysis on a 12 megapixel TIFF on a dual-core 2.93Ghz E6500 under Win7 64bit
• File browser doesn't remember last location
• There is no way to zoom in on the picture AFAICT
• No way to enter numerical adjustments, using the sliders is fiddly if a certain exact figure is needed
• A visual hue gauge would be helpful in the H mode under editing controls to determine where one needs to move the slider to get to a certain color
• The maps are a nice touch, and clicking on the color to show where each is mapped on the image is even nicer
• The UI is quite good, but the Editing Controls are not very intuitive and un-documented (more informative tooltips would be helpful)

Questions:

• Is the HSV conversion destructive?
• Do you have plans to implement this as a plugin for Lightroom? I'd pay good money for that
• Is it really fully 16-bit under the hood?
• Is it ICC profile aware? The ProPhoto TIFFs I exported from LR weren't color managed properly in Phozn
• What's the function of Balance sliders?
• What happens when I move a slider up or down - does it nudge the adjacent clusters up/down, or is there a threat of clipping?

Was playing around with the attached photo. I changed the hue on the sisters' dresses which worked nicely - they were originally teal on the left and light green on the right -, and increased the Value Intensity of skin tones slightly (56-63, see screencap). Results had bad posterization in the in-program review, but the hues were right. Exported JPEG attached - one might note some slight hue shifts on skin tones. A bug or did I do something wrong?

[Mark D Segal]

Mark,

thanks for your comments. This is why I proposed my application to a community of experts as this forum is.

I'm not a legal expert so I used a quite standard license agreement. Do you have an idea of where I can find some conditions that would better fit your expectations?

As I wrote in the download section, if anyone is interested in testing the application without restrictions he has just to get in touch. The only restriction is in the size (not the format) of the saved images, tough.  I'm planning to end this Beta phase at the end of March 2011. I still don't have a commercial plan after that.

I can assure there is no harm to your operating system in using Phozn. It is developed in Java and among the benefits there is the fact that it runs in a sort of "sandbox" (i.e. the Java Virtual Machine).

Phozn can get in input 8/16 bit TIFF or 8 bit JPEG images. It can produce in output the very same formats, so you can have a completely 16 bit Digital Workflow using the TIFF file format.

It is a standalone application and one of the best use case is for a final tuning after a RAW converter. But it can be used after a more complex Photoshop-based editing as well. An output image can be used as a layer in Photoshop and mixed with the original image to create more complex effects. I'm planning a tighter integration with Photoshop and Lightroom in the future but I'm in an early stage right now.

What is different in respect to other image editors is the ability to automatically segment an image in different sets (or zones, if you prefer) in respect to Hue, Saturation or Value of the pixels (as per the HSV Color Model). E.g. in the tutorial I pointed in my previous message you can "select" the sky from the rest of the picture just asking to run the "cluster analysis" engine on hue with 2 sets (one will be the sky, one will be everything else).

With 3 sets you could easily segment an image between the blues (sky, water), the greens and the browns. And so on. There is currently no concept of "locality" in the segmentation algorithms right now, but I'm thinking about that.

I know Phozn is not very user friendly right know, this is why I'm asking for some feedbacks (on the functionalities, on the user interface, etc).

Thanks again for your contributions!

Rgrds,
Danilo.

Danilo, at this stage it isn't a matter of legal stuff, but more what is useful to people testing an application. If your response to the size issue is "tough", well so be it, but I won't spend time with it unless I can play with and save full-sized, full-featured images this application is capable of producing. Sorry, but time is short and I can't be spending it to try seriously evaluating crippled versions of applications.

[danilop]

Mark,

if you are (even slightly) interested in an "uncrippled" version of my application I'd be more than glad to give you a registration key for the beta period. I have made no plans for what's next, but whoever will spend some time testing Phozn will have my sincere gratitude at least.

The registration key is bound to an email address, so if you'd like to have it please send me a Personal Message. If you don't feel comfortable with this procedure please tell me how you'd manage this in a better way. I'm ready for changes!

Rgrds,
Danilo.

[degrub]

Mark,

Perhaps there was a typo in his response. The word "though" would make more sense in context, especially since he has not shown such hostility in the rest of the correspondence. Both are valid words in spell check. How many times have we been caught on that one   !

Whether it is worth spending time testing at this point is hard to say. Each to his own judgment.

Frank

[danilop]

Hi feppe,

thanks for your comments!

45 minuted for a 12MP TIFF is quite long, I agree!

I'm developing on a Mac and I run cluster analysis in a couple of minutes on my 18MP 16-bit TIFFs (produced by Canon DPP out of my 550D RAW files). I just did some tests on a Windows box (32-bit Vista) using 12MP JPEGs (produced by a Canon S90) and the time is nearly the same.

How many sets did you ask for? The convergence time for the cluster analysis algorithm grows quickly with the number of sets. Moreover, I found that a number of sets higher than 4-5 will often (almost always) produce zones that are not quite visually separate from each other and so different tuning cannot be applied without creating some "noise" in the resulting image.

My suggestion is to ask for 2-3 sets and eventually run more cluster analysis with different parameters (and different tuning, of course).

The only way to "zoom" right now is to click on the image in the central panel (usually the preview image): this will open a new window containing all the image and such a window can be maximized up to the screen size. You can't see the pixel details in this way, tough.

I'll put this items on my todo list (thanks for the suggestions!):
- No way to enter numerical adjustments, using the sliders is fiddly if a certain exact figure is needed
- A visual hue gauge would be helpful in the H mode under editing controls to determine where one needs to move the slider to get to a certain color
- The maps are a nice touch, and clicking on the color to show where each is mapped on the image is even nicer
- The UI is quite good, but the Editing Controls are not very intuitive and un-documented (more informative tooltips would be helpful)

Going to the questions:
- Is the HSV conversion destructive?
I'm moving 16 bit integer data to a (currently) 32 bit floating point, so I don't think there is any loss of precision in the conversion. I did some tests and going back and forward to HSV with my implementation gives the very same 16-bit original RGB data.
- Do you have plans to implement this as a plugin for Lightroom? I'd pay good money for that.
I have some plans, but I'd like to get a stable standalone application at first.
- Is it really fully 16-bit under the hood?
Yes, absolutely! It was my requirement as I'm using Phozn for the final fine tunings after my RAW converter. You can see that by the fact that if you use a 16-bit (TIFF) file in input you don't get visual "hole" in the histograms. Please compare that to a 8-bit input file (e.g. a JPEG)
- Is it ICC profile aware? The ProPhoto TIFFs I exported from LR weren't color managed properly in Phozn
Unfortunately not. I have to work on that!
- What's the function of Balance sliders?
The "Intensity" sliders increase or decrease all the values they affect by a certain factor. If Si is < 50 (e.g. 40) all Saturation values are multiplied by Si/50 (e.g by 40/50=0.8 ), if Si is > 50 (e.g. 60) it's the "opposite" but, to avoid clipping, I use the following formula: 100 - (100 - current value) * (100 - Si) / 50 (at first it is complicated but at a second look it makes sense).
The "Balance" sliders will "compress" one half of the values and expand the other half. E.g. if I set Sb to 40 it will compress Saturation values in the range 0-50 (50 is "half") to 0-40 and expand Saturation values in the range 50-100 to 40-100. In this way "Balance" is more smooth/soft on the edges (0 and 100) and more strong for "middle" values.
- What happens when I move a slider up or down - does it nudge the adjacent clusters up/down, or is there a threat of clipping?
As per the above formulas there is no clipping.

I'll investigate the posterization on the preview window you mentioned.

Thanks again for your efforts!

Rgrds,
Danilo.

[Mark D Segal]

Mark,

Perhaps there was a typo in his response. The word "though" would make more sense in context, especially since he has not shown such hostility in the rest of the correspondence. Both are valid words in spell check. How many times have we been caught on that one   !

Whether it is worth spending time testing at this point is hard to say. Each to his own judgment.

Frank

Frank - very astute of you - that didn't occur to me. Being kind of busy I just read stuff as it comes across and respond accordingly, but I think you may well be right. Danilo seems to be accommodating and his entrepreneurship should be encouraged.

[feppe]

You're welcome, I'm glad I could be of help!

45 minuted for a 12MP TIFF is quite long, I agree!

I'm developing on a Mac and I run cluster analysis in a couple of minutes on my 18MP 16-bit TIFFs (produced by Canon DPP out of my 550D RAW files). I just did some tests on a Windows box (32-bit Vista) using 12MP JPEGs (produced by a Canon S90) and the time is nearly the same.

How many sets did you ask for? The convergence time for the cluster analysis algorithm grows quickly with the number of sets. Moreover, I found that a number of sets higher than 4-5 will often (almost always) produce zones that are not quite visually separate from each other and so different tuning cannot be applied without creating some "noise" in the resulting image.

My suggestion is to ask for 2-3 sets and eventually run more cluster analysis with different parameters (and different tuning, of course).

I went for 11 sets straight away, so that would explain it.

I'll investigate the posterization on the preview window you mentioned.

It's not the posterization in the preview I'm too concerned about, but how the skin highlights took a hue in the final output JPEG, hue which is not visible in the preview. I didn't have any hue adjustment on the skin tones.

[Drazick]

Hello,
Any updates on the program?