Questions on next version 5.6

[JPD]

Dave I saw your two files.

What I must say to be honnest is that the resizing algorithm of PTE is very good, the problem, enough simple to solve is in logical built between screen and format. For the 3 differents mode there must be a good solution (see post #158), I am working to make it easy to understand, I haven't yet finish.

[CorVdK]

Hello Dave,

For your info.

I have seen your demo experiment.

TRACTOR 1 = I have seen it exactly as you describe it. Very smooth on my computer (Quad CPU, 3 GB RAM, NVidia Geforce 8600 GT 512 Mb).

TRACTOR 2 - TV SAFE ZONE = Also very smooth, 10 % TV zone: OK. But as described by you, 1-pixel lines between the black strips. Also the rectangles on the random basis, are showing a "raster" when they come together. Disappearing after +/- 1 second.

Cor

[xahu34]

Hello Dave,

If you parquet your screen with butted stripes or rectangles, and want to resize the whole stuff, you can go two ways:

1. You can combine the objects to one single image before resizing.

2. You can resize each object separately before putting everything together.

The first way will lead to rather nice results. If you have to go the second way, the effect (thin lines) which you could see, is quite natural. The screen is not a continuum; you have to round positions and sizes.

Best regards,

Xaver

[JPD]

However, things now begin to fall apart and thin lines appear between the various strips and rectangles on a random basis.

This led me to go back to the TRACTOR 1 project and try zooming the parent image while the assembly of the strips was taking place. Sure enough, the random lines began to appear. They also appear in the third slide where the 25 rectangles are zoomed and rotated.

I often had this problem and see that also one some slideshows of friends, it often depend of the graphic card we use. An example is the slideshow of a Diapositif members here which have such problem. The topic can be see only by Diapositif's member here but what said people :

Without default :

Radéon 9250 en 1600 x 1200

nVidia GeForce 7900 GS Sortie DVI écran 23" 1680x1050

Radéon 9250 en 1280x960

With default :

NVIDIA GeForce 8600M GT GPU 512 Mo de ram Résolution 1920 x 1200

NVIDIA Ge Force 8600GT 512 Mo Sortie DVI écran 1920x1200

Radeon 9200 en 128 Mo

nvidia gforce fx 5500 et 256 mo. écran 24 pouces.

I made measurement and it seems Igor used the second explanation of Xaver in order to resize only once each object for best final quality (it will not for mask I think but haven't yet make measurement, so not sure). I think Igor is right to have made this choice. This little problem as often a solution, put one pixel more on each rectangle. I don't think it's a problem of quality of the resizing which is very good if you remember it is abble to resize hundred of objects more than 50/60 times each second even on old graphic card (for example Fantaisies florales)... a great job of Wnsoft.

[xahu34]

Hello Dave,

Would it be possible for you to provide new versions of slides 1 and 2 of your tractor sequences, by replacing stripes 1 to 9 by new ones whose width is enlarged by 1 pixel on the right side? This would mean that all adjacent stripes would have a 1-pixel overlap (see the remark of JPD from above). I just had a look at one of JPD's constructions built with the cale-method. He seems to apply this technique.

Best regards,

Xaver

[davegee]

Hello Xaver,

I have sent you a personal message.

DaveG

[JPD]

Jean-Pierre,

I have posted the results of my tests here:

http://www.mediafire.com/?sharekey=411ad3b...3a805876665040c

May I say that tables of mathematical data did not convince me that there was a problem – I needed to see the problem with my own eyes with actual images. Manufactured demonstrations using rectangles are also unconvincing. So I set about constructing a test using actual images at "Actual Pixels" (equivalent to Original Mode) to see for myself what the problems are (if any).

DaveG

Dave,

Here are 2 tests with Cale method I did with the template you send me, one is 100% the other 90%. as you can see, it's exactly what you did, I just modify two files in order to have no default, the weight is only 2 722 Kb, I used only JPG, vs 11 715 Kb for yours (made with V5.52).

There were problem with your fullscreen test at 1024 x 768 and 1152 x 864 screen definition and for your 90% screen test a problem at the same screen definition plus at 1280 x 960 (1024)

The two tests I did have no problem from 800 x 600 to 2048 x 1536, I have test them on 10 screen definition. I haven't use 1 pixel more pictures but another way to correct the problem of black lines.

As you also can see, the 90% screen version haven't the problem of objects outside.

What I did would be possible with Fit to slide screen because it's only 2 levels project, but would be more difficult to do than with Cale method, it would be so easy to do as Cale method with the sugestions I did.

I will put the templates if people are interested to see how I did.

[davegee]

JP,

I'm sure that everyone would like to see how the Cale method works on actual images. So yes please, will you post the templates?

To clarify, these modified tests were done using 5.52?

On the subject of the problems you saw in my full screen test at 1024x768.

Could it be because I used monitors whose NATIVE RESOLUTION is 1024x768 and 1280x1024 whereas you are using the same 2048x1536 monitor to SIMULATE these resolutions?

In my opinion, this is not the same thing.

Tractor 1 works perfectly on monitors whose Native Resolution is 1920x1200, 1280x1024 and 1024x768. I only ever use Native Resolution on all of my monitors (3).

DaveG

[JPD]

JP,

I'm sure that everyone would like to see how the Cale method works on actual images. So yes please, will you post the templates?

It's here with both 5.52 and 5.6 templates

To clarify, these modified tests were done using 5.52?

The 2 examples are with 5.52, 5.6 hasn't original mode, the 5.6 template read only the 5.52 template and I add 2 masks

On the subject of the problems you saw in my full screen test at 1024x768.

Could it be because I used monitors whose NATIVE RESOLUTION is 1024x768 and 1280x1024 whereas you are using the same 2048x1536 monitor to SIMULATE these resolutions?

In my opinion, this is not the same thing.

Tractor 1 works perfectly on monitors whose Native Resolution is 1920x1200, 1280x1024 and 1024x768. I only ever use Native Resolution on all of my monitors (3).

As I wrote before it depend of graphic card, with mine there are the problem I wrote.

DaveG

What would nice is to do the same with V5.6, without needing a Cale and using original mode and position in Pixel without needing a Size/position box.

[JPD]

Hello Jean-Pierre,

let me make a final remark on your solution. For me, PTE is a black box. I know its interfaces, how to insert images, how to configure transitions, how to set animation parameters. Furthermore I can look at PTE's output, the preview or an exe-file (I do now use the video stuff). This is what I see, and that's all. I think that your solution would have substantial effects on the internal structure of the PTE software, which is invisible for me as a user. Even if I could look at the design specifications and the source code, this would not help. I am not a software developer with experience in graphics applications.

So I do not feel myself in the position to make any substantial remark on your solution. I'm sorry about that, and I would very glad if future versions of PTE would serve your demands.

Best regards,

Xaver

I have prepared this explanation for the following purposes:

- to give to Igor and the team at WnSoft the details of the formulas that they will need in order to program my algorithm into PTE

- to give to those who currently use Original mode, reassurance that this new method will give the same end result as now

- to give to those who currently use other modes, reassurance that this new method will give the same end result as now

I have made available the following material:

- an Excel spreadsheet file containing the formulas (link here)

- a gif file that is an image of the spreadsheet for those who would like to see the spreadsheet but don't have Excel (link here)

- a zipped PTE project file that demonstrates the result of using v5.52 as we do today and the same result achieved using my algorithm (link here)

- an image made up of some screen-scrapes that shows which fields have been used in the O&A window (see below)

Test of the process :

The PTE project contains pairs of images (#2 and #4, #6 and #8, #10 and #12). Slides #2, #6 and #10 are the ones set up using v5.52 features as we do today. Slides #4, #8 and #12 are the ones set up using my algorithm. This was achieved by entering into the spreadsheet the appropriate values (the same values that we would have entered via PTE if it had the algorithm already built in). I then took the calculated results and keyed those numbers into those fields which are available today in v5.52.

Yes, it is only a simulation of the algorithm but it is, I believe, a very accurate simulation.

Slide #2 demonstrates the result for objects that use "Original mode" as many people use with v5.52 today. Slide #6 demonstrates the result on objects of all three of the modes: "Fit to Slide", "Original" and "Cover slide". Slide #10 demonstrates the result on a rotated object (one object from each of the three modes).

The real test comes from slide #13 onwards. Here I show the pairs of slides, one after the other. The only way you should know when the slide has changed is to watch the number at the top of the screen.

In order not to confuse you with too much information, my examples all have the Level1 object as "Fit to Slide". I have two other spreadsheets that handle the situation when Level1 is "Original" and when Level1 is "Cover slide". I can make these available to anyone who wishes to study them. And I will obviously make them available to Igor and his team.

Using this algorithm and with the virtual Format as parent as explained in post #82, there will be no need for anyone to change anything – so all users should be happy. And there will be only one algorithm to code now and to test in future - which should make everyone at WnSoft happy. Also, we no longer need the "Size/Position…" window! Finally, with this solution, we no longer need point 5 of post #82. Simpler for everybody!!!!

Thanks to Peter to help me for translation

[Igor]

Thanks, I'm working with provided test projects and calculations to understand the problem.

[JPD]

Thanks, I'm working with provided test projects and calculations to understand the problem.

I just send you a word document in which are all the differents posts with suggestions, it correct some suggestions we don't need since it' possible to have the 3 modes, the document is more homogen and easiest to use for you.

[JPD]

Here is the link to the word document about "3 modes as V5.52 with only one algorithm as V5.6" I send 5 days ago to Wnsoft, I always hope it will be useful for Wnsoft and for PTE users.

[Hartmut Scherer]

JP,

Did you hear anything regarding your proposal?

[JPD]

JP,

Did you hear anything regarding your proposal?

Igor wrote me about it, but to day I am not sure that my proposal have been understood, he is always working on his Size position box, I am very pessimist and tired.

Many people seems not understand that we have finally a screen format very interesting for slideshow which is avalaible for many years, at least 10 or 20 years, the full HD format and we can use it without any resizing inside PTE if we use the original mode (for nominal format of course).

When you put the position of a picture in percent you aren't abble to say if each pixel of the picture if not at 0% of pan is exactly on one pixel of the screen, if not your picture's pixel will be calculated on 2 or 4 screen's pixels with a lose of quality. It's the same thing when you use Fit to slide or Cover slide, you are enabble to say if your picture is resized or not when it's on level 2, 3, 4 or more level.

Why have soo perfect camera as we have to day and add this default.

PTE, since the beginning had this wonderfull possibility that others haven't, it's now we are losing this quality, just at the moment we can use it better than never.