YOUR ACCOUNT

Login or Register to post new topics or replies
miome
Posts: 7
Hello. I have a very specific problem I am trying to solve. I was hoping some one here might know the answer. (Techniques, software, any knowledge would be appreciated)

I have several pictures that I have applied filters to in Photoshop 6.0. I originally scanned/rendered these images @ 800 x 600 (72 dpi) so that I could try out various filters and multiple combinations of filters without waiting a long time for each filter to finish. I was planning on figuring out which filters and filter settings I wanted to use, then importing the source images into photoshop at 3200 x 2400 (and some as large as 9600 x 7200) to apply the combinations of filters I had decided to use.

The problem is that the resolution of the filters stays the same even though the image size has increased. This results in the filters looking quite a bit different when applied to a large file than they do when applied to an 800 x 600 file. The settings in the photoshop filters are already maxed so I cant compensate by changing those values.

So ultimately what I would like to know is:

A: is there some kind of filter preferences setting or some other file size setting I missed that will allow me to counter act this phenomenon?

B: Does photoshop CS2 resolve this issue? If so, does CS2 have all the exact same filters because at this point the compositions are finished. What I want to do is start with the large files and run the exact same filters I used on the 800 x 600 files resulting in 3200 x 2400 filtered images that look identical to my 800 x 600 filtered images.

C: Are there any plugins, updates for 6.0, or third party software that have addressed this issue?

D: Is there a way to start with my large files, reduce the pixel count to 800 x 600 while increasing the DPI resulting in a smaller image size with no information loss, then run the photoshop filters, then blow the pictures backup up while reducing the DPI so that the entire process would remain lossless? (this one is wishful I know! I kind of asked it just for kicks. smile:-)

E: Has anyone tried printing a 300 dpi image at 2 x 2 inches then enlarging the picture photographically to 3 x 2 feet? I know there is new fractal enlargement software out there but I think telling the computer to interpolate that much information is probably pushing it.

I found this website because of a Google search that yielded "resolution independant" filters so I figured there would be people here that are aware of this issue. The amazing thing is that so many of the filters in photoshop 6.0 actually seem to work/look best at 800 x 600. The canvas texture for example actually tiles at anything larger than 800 x 600 resutling in an image that looks like it has 100 identical canvas chunks in it. Not that I would use this filter, it's just a prime example of the fact that the filters don't seem to have been designed for higher resolutions.

By the way Filter Forge looks awesome, I plan on getting it! smile:-)

Thanks for any help.

One humble newbie,

Miome.
  Details E-Mail
Crapadilla
lvl 52 Filter Weaver and Official "Filter Forge Seer"

Posts: 4365
Filters: 65
Quote
miome wrote:
The problem is that the resolution of the filters stays the same even though the image size has increased. This results in the filters looking quite a bit different when applied to a large file than they do when applied to an 800 x 600 file.


Technically, there is no such thing as a 'resolution' of a filter. FF has a sample-based architecture, and with larger images the rendering engine simply makes more evaluations of the filter's image function at any given area of the image. As such, FF is resolution independent as advertised.

However - internally - many filters are 'hardwired' to a certain maximum detail level for the sake of rendering performance: The high detail that would be possible is not required for low-resolution renderings (most filters are 'optimized' for the filter library resolution of 600x600 px). This internal 'detail restriction' is often done with filters that use noise components, since these can be very slow. Also, filters that use blur-type components may look different at higher resolutions because the blur radius is measured as a percentage of image size.

Still, it would probably help if you posted some examples from those filters that supposedly look different at high resolutions. After all, the filter authors might be willing to update these to 'work' as expected with higher resolutions... smile;)
--- Crapadilla says: "Damn you, stupid redundant feature requests!" ;)
  Details E-Mail
miome
Posts: 7
Hi thanks for the response. I don't have any examples with me that I can upload.

Incidentally I dont need the filter detail to go higher, I need the detail setting to go lower. Using the chrome filter with the detail slider set as low as it can go, and the smoothness filter set as high as it can go, I achieve a chrome reflection pattern on the 800 x 600 files that have a certain amount of space around and between the reflection lines that filter produces.

When I start with a much larger file, then run the same filter and settings I did on the 800x 600 files, the reflection patterns hug the lines of the picture/art much more tightly resulting in reflection patterns that are much more detailed and numerous than I want. Since I already started with the detail slider set to it's lowest value, I can't use this to compensate.

I think CS2 may have solved this issue by designing the filter setting sliders to go much higher and lower, this would allow me to achive the same detail/image size ratio on a large file that I achieve on a small file. I just don't have access to it to find out before buying it.

I have found the same issue to affect the water color, and dry brush filters. Imagine the "brush size" setting is set to approximate the size of a dime on an 8.5 x 11 inch sheet of paper. The paper has a daisy drawn on it. The brush "blots" which were the size of a dime would align themselves adjacently around the lines of the daisy.

Now, start with a piece of paper the size of a 3 x 4 foot poster. The poster has a much larger daisy on it because it is just an enlarged copy of the 8.5 x 11 image.

The problem is when I run the filter on the 3 x 4 or poster size image, the brush size (or "blots") is still the size of a dime. So now many more dime size blots can fit around the outline of the daisy, making the brush "blots" smaller and more numerous in relation to the daisy then when the same filter is run on the 8.5 x 11 image.

If I could increase the filter brush size so that it was the size of a quarter (give or take) and then use the quarter size "brush blots" on the larger image, then I could achive the same "brush blot" size in relation to the daisy.

I have found this to phenomenon to affect all the filters I have tried in photoshop 6.0 including the chrome, plastic wrap, water color, dry brush, and texture settings such as brick, canvas, craqelure etc. ("detail" settings for chrome and plastic wrap being the equivilant of "brush size" for dry brush etc.)

In the future I plan to use filters that are "resolution independant" from the start so I wont wind up with this problem. In the mean time I have several compositions that have combinations of filters, and fade/overlay settings that were rendered in photoshop. It would be impossible to make the compositions look the same in any other application.

Thanks very much for your help.

-Miome

  Details E-Mail
Crapadilla
lvl 52 Filter Weaver and Official "Filter Forge Seer"

Posts: 4365
Filters: 65
Ah, there was a misunderstanding. You are you talking about Photoshop's very own built-in filters, not the filters supplied by the FF plugin, right?
--- Crapadilla says: "Damn you, stupid redundant feature requests!" ;)
  Details E-Mail
miome
Posts: 7
I apologize for the confusion. Yes, I was talking about photoshop's built in filters.

Shoot, when you said the filter's "authors" might be willing to fix them, I got the impression there were people here who had actually written photoshp's filters. That would have been handy! smile:-)

Are you or is any one familiar with this diemna as it ralates specifically to photoshop? How did people deal with this? Are there any conversion processes or updates that fix photoshop's filters in this area? And also does any one have cs2 and know if this is resolved? I would be willing to purchase the new photoshop if I knew I could adjust the filter parameters to make the filters look the same on the large files as they do on the small files.

Thanks again for your time, and I apologize for not being more specific. I can totally see where people would think I was talking about Filter Forge and I wasnt! smile:)

From what I can see Filter Forge looks like an AMAZING collection of filters.

I hope I didnt put any one off from helping me! smile:cry:

miome smile:)
  Details E-Mail
Sphinx.
Filter Optimizer

Posts: 1750
Filters: 39
I've "invented" a small trick you can use to work around such restrictions, however its a crude post processing estimation that only works in some cases. It was designed to handle the opposite problem of what you describe (i.e. reducing a filter to subpixel processing), but read on - it may help you here too.

The trick allows you to resample only the effects of a filter (i.e. what a filter changed). It works good when you want a filter to work at subpixel level (i.e. reducing the spread of a filter). When you use the trick to gain spatially wider effects than what the given filter allows, the effect changes are resampled, which can ruin details and introduce resampling artifacts. However if the filter does not produce new high frequencies it could work (blur filters etc. are good here).

Here is how it works:

1. Create a copy of your original image and resample it down by a factor equal to the spatial spread you want to increase in the filter. E.g. if you want the filter to have the double radius, you resample the copy down to 50% (if you wanted to reduce the spread of the filter by a factor 2 you'd scale up to 200% instead)

2. Create a copy of the background layer (so you have two similar downsampled layers).

3. Apply the filter(s) you have in mind to the new layer. IMPORTANT! If your filters support 16 bit per channel ("image -> mode"), you should change the image mode to 16 bit per channel before starting the filtering.

4. Invert the filtered layer and set its opacity to 50%. If you couldn't set the image mode to 16 bit per channel like described above, you should do it now. What you see now is a gray weird looking mix - don't worry about that (it's what we want).

5. Merge down the filtered layer so you only have one layer (the weird gray version), invert this layer, and resample the image back to the original size.

6. Switch to the original image and set it to 16 bit per channel. Switch to the filtered image again (the gray weird looking one), drag over the layer to your original image (make sure you align the layer precisely), and set that new layers blendmode to "Linear Light".

Now you see the effect of the filter overlayed perfectly on your original resolution image - however you'll most probably notice some bad looking resampling artifacts - one way to reduce these is to gently apply gaussian blur to that layer (small radius, e.g. 1)

When you are satisfied with the effect blend, you merge down the layer and eventually set the image mode back to 8 bits per channel.

If your process is composed of several filters, you should try to restrict the above workaround to only the given filter step you find restricting.


Hope it helps,

Regards,

Michael
  Details E-Mail
Bella
Moderator
Filter Forge, Inc.
Posts: 274
Miome,

I'm moving this thread to the Offtopic forum. General Discussion is supposed to be for Filter-Forge-related topics.
  Details E-Mail
miome
Posts: 7
Bella, my bad. I didn't mean to post incorrectly.

I think I must have thought "off topic" meant like politics and stuff. I didnt bother to even look in the forum to see what was in there. smile:?:

Sorry, -Miome.
  Details E-Mail
miome
Posts: 7
Micheal. Thanks very much for the detailed response. I really appreciate the knowledge and effort to help. My problem is some of the compositions invlove several filters combined. I can't imagine doing that to all of them and not having the artifacts start to compound.

I downloaded CS3 trial version and they STILL haven't addressed this issue! I am blown away by the fact that you cant run a filter on an 800 x 600 image then run that same filter on a 3200 x 2400 version of the image and yield identical results. All they would have to do is add a "process by percent of image size" option to the filter parameter sliders and that would solve the problem.

The main thing is the filters don't actually look good when run on large files, or even function as intended so basically all the photoshop filters (for effects I mean) are optimized for low resolution.

I am looking now instead at printing 300 dpi at 8.5 x 11 then re-scanning at a high resolution to get my 2 x 3 feet and just living with the blown up printer dots. Of course there will be two sets of printer dots on the final image, I don't know how freakish that will be. (probably not glicee material)

Thanks againI really appreciate your detailed response.

-Miome
  Details E-Mail
Sphinx.
Filter Optimizer

Posts: 1750
Filters: 39
Miome,

Could you post a sample image of your low resolution filtering (before and after)? The print thing you talk about will really just give you an effect of resampling + diffusion, you could easily do that in photoshop without printing...
  Details E-Mail
miome
Posts: 7
Micheal, I cant upload an example right now because my art is in a seperate location. It's very easy to duplicate this phenominon. Just scan a picture at 3200 x 2400 and then again at 800 x 600. Run the chrome or dry brush filter on both with the same settings and they will look different. Or you can render something (vector) at both pixel image sizes and run a filter on the different size images, that works also.

When you say I could enlarge in photoshop do you mean use resampling? Because I have tried that and it produces pixelated edges. I think I would rather live with a rescanned image or a photographically enlarged image.

I haven't tried any fractal enlargement software yet, but I am not totally opposed to it. Especially if I could enlarge just the filter effects and overlay them on the image.

Thanks,
-Miome
  Details E-Mail
Sphinx.
Filter Optimizer

Posts: 1750
Filters: 39
you should try out the method I mention, just for a test case. Its only the effect you'll overlay..so its only the effect itself that will suffer, not the underlying image. Its important that you use bicubic or similar resampling when you sample up the image again.
  Details E-Mail

Join Our Community!

Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!

33,711 Registered Users
+18 new in 30 days!

153,533 Posts
+38 new in 30 days!

15,348 Topics
+73 new in year!

Create an Account

Online Users Last minute:

28 unregistered users.

Recent Forum Posts: