Re: [Gimp-developer] index colour images: interp
Quoting [EMAIL PROTECTED]: Once 2.4 is out and there is a review of the interp naming strategy w.r.t. downscaling the use of NONE should be probably be changed as well. (A scaled up image with no interpolation has holes in it.) Even though rather simplistic, N.N _is_ interpolation. ;) It is not a hard-and-fast rule that the label of a menu item has to precisely and accurately describe the functionality of that menu item; indeed, it is sometimes preferable to address the general concept being addressed by the item, realizing that those aware of the technical inaccuracies would not be confused by the mislabeling, while those who are not might likely be confused by an accurate label (or in consideration of a fairly reasonable association which contributes to a simplified UI). Case in point: most word processors have an Insert footer command while most typesetters would argue that you don't insert a footer, you attach a footer. It is overly presumptuous, in my opinion, to declare that labeling nearest neighbor interpolation as None is an error on the part of the user interface designers of the GIMP. It may be worthwhile to propose re-examination of the best approach, but it should not be assumed that the existing labeling is owing to any lack of consideration by the developers. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Sat, 23 Jun 2007 08:53:16 +0200, [EMAIL PROTECTED] wrote: It is overly presumptuous, in my opinion, to declare that labeling nearest neighbor interpolation as None is an error on the part of the user interface designers of the GIMP. It may be worthwhile to propose re-examination of the best approach, but it should not be assumed that the existing labeling is owing to any lack of consideration by the developers. Indexed colour layers are always scaled without interpolation. This is incorrect. I suggest the following. Indexed colour layers are always scaled using nearest neighbour interpolation. So I did not say it was an error due to neglect. I said it was incorrect. That is a FACT. As I pointed out N.N. _is_ interpolation. Once 2.4 is out and there is a review of the interp naming strategy w.r.t. downscalingthe use of NONE should be probably be changed as well. Note the conditional should and the uncertainty probably. I'm not being dogmatic or presumptuous. I'm doing _exactly_ what you suggest proposing a re-examination and also proposing an alternative that I consider better. If you cant hack a polite , wieghed critism with a suggested improvment without misinterpreting it as an attack on your work I suggest stop reading gimp-devel which is where such critisims should be posted. Please bear in mind the common aim is to improve Gimp. Case in point: most word processors have an Insert footer command while most typesetters would argue that you don't insert a footer, you attach a footer. So MS did one other their typical redefinitions, everyones mindlessly copies their mistake because Windows is GOD and now you prospose this a model behaviour as if it somehow backs up your case for an incorrect label in Gimp. You may of may not have a point about simplicity but personnally I dont like what is, yet again, the MS approach of dumbing down the user. Gimp is not aimed at a moronic click and share home user. I think it is important to be accurate and NONE is wrong. There maybe other options but any detailed discussion should probably wait until the whole issue is reviewed after 2.4 release. Thanks for you comments. /gg ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
Hi, On Sat, 2007-06-23 at 10:44 +0200, [EMAIL PROTECTED] wrote: Note the conditional should and the uncertainty probably. I'm not being dogmatic or presumptuous. I'm doing _exactly_ what you suggest proposing a re-examination and also proposing an alternative that I consider better. That's perfectly valid and we may even reconsider it. But then it used to be called nearest-neighbour and at some point we spent a lot of thought and discussion on these labels and changed them to what we have now. Somehow I am not very inclined to have this discussion again. It seems like a waste of time to change things back and forth when we could spend that time to move forward instead. Sven ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Sat, 23 Jun 2007 11:57:23 +0200, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Sat, 2007-06-23 at 10:44 +0200, [EMAIL PROTECTED] wrote: Note the conditional should and the uncertainty probably. I'm not being dogmatic or presumptuous. I'm doing _exactly_ what you suggest proposing a re-examination and also proposing an alternative that I consider better. That's perfectly valid and we may even reconsider it. But then it used to be called nearest-neighbour and at some point we spent a lot of thought and discussion on these labels and changed them to what we have now. Somehow I am not very inclined to have this discussion again. It seems like a waste of time to change things back and forth when we could spend that time to move forward instead. Sven Thanks, indeed it should move forwards rather than oscillate. Like I said in the relevant bug report there will be a need to re-evaluate all of these labels in relation to interpolate/decimate which is actually quite different code. NONE would clearly be part of that discussion. However this has diviated from my suggestion which was nothing to do with the NONE label but rather with indexed pallette warning. I think nearest neighbour is non technical, very obvious in it's meaning and readily understood. I dont see the sense in Gimp stating that it does something it does not in a warning that is supposed to clearify what happens. I suggeste: c/Indexed colour layers are always scaled without interpolation/Indexed colour layers are always scaled using basic nearest neighbour interpolation/ over and out. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
Hi, On Sat, 2007-06-23 at 13:26 +0200, [EMAIL PROTECTED] wrote: I think nearest neighbour is non technical, very obvious in it's meaning and readily understood. IMO it is very technical and the vast majority of users does not understand its meaning. They also don't understand Linear or Cubic for that matter, but it's difficult to come up with simpler terms. I dont see the sense in Gimp stating that it does something it does not in a warning that is supposed to clearify what happens. I suggeste: c/Indexed colour layers are always scaled without interpolation/Indexed colour layers are always scaled using basic nearest neighbour interpolation/ If we use None in the combo-box, then we also have to use without interpolation in the text below it. If we changed this text to use nearest-neighbour we would also have to use that term in the combo-box. Since we postponed that change for now, your suggestion is rejected. Sven ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
David Gowers wrote: On 6/23/07, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Sat, 2007-06-23 at 13:26 +0200, [EMAIL PROTECTED] wrote: I think nearest neighbour is non technical, very obvious in it's meaning and readily understood. IMO it is very technical and the vast majority of users does not understand its meaning. They also don't understand Linear or Cubic for that matter, but it's difficult to come up with simpler terms. Well, we could consider naming it in a slightly longer way but one that may be more suggestive to the ordinary user --- 1-factor (nearest neighbour) 2-factor (linear) 3-factor (cubic) 5-factor (lanczos3) -- is this number correct? So it would be a fairly simple and common pattern to see -- more factors taken into consideration - more quality, more time. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer Couldn't we suggest in there naming what is the most appropriate filter for a certain image type. linear = line-art (low) nearest neighbour = line-art (medium) cubic = photo(high/medium) lanczos = photo(high) Geert ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Sat, 23 Jun 2007 16:34:28 +0200, Geert Jordaens [EMAIL PROTECTED] wrote: David Gowers wrote: On 6/23/07, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Sat, 2007-06-23 at 13:26 +0200, [EMAIL PROTECTED] wrote: I think nearest neighbour is non technical, very obvious in it's meaning and readily understood. IMO it is very technical and the vast majority of users does not understand its meaning. They also don't understand Linear or Cubic for that matter, but it's difficult to come up with simpler terms. Well, we could consider naming it in a slightly longer way but one that may be more suggestive to the ordinary user --- 1-factor (nearest neighbour) 2-factor (linear) 3-factor (cubic) 5-factor (lanczos3) -- is this number correct? So it would be a fairly simple and common pattern to see -- more factors taken into consideration - more quality, more time. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer Couldn't we suggest in there naming what is the most appropriate filter for a certain image type. linear = line-art (low) nearest neighbour = line-art (medium) cubic = photo(high/medium) lanczos = photo(high) Geert OK, looks like we get a pre-discussion now, although be warned implementation has been put off to 2.4 Sven , I can understand someone without a knowlege of maths not understanding what linear and cubic refer to , they'll have to suck-and-see or read the doc. but anyone capable walking to the shop for a loaf of bread ought to be able to work out what nearest nieghbour means and have a fairly good guess as to what it means in pixels. David , yes lanczos3 is pretty sure to be the most suitable option. The code is now general enought to swap to lanczos4 with c/3/4/ in one const definition but it's not so good a v. slow. Geert, linear is way better than N.N. you need to swap that around. I'd like to see some extra info available like a tool tip but I dont think the currect dropdown list object has that behaviour. Could be wrong tho'. I think this whole subject needs going into thoroughly especially w.r.t. downscaling where decimation is used in place of interp. It would be good for Peter to come in and cast his expert eye on whether this could better be presented in different way in the UI. But as this will not be a trivial discussion the descission was to put all this off 'till later and probably tackle it as part of GEGL integration. Thanks for your thoughts. /gg ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Fri, 22 Jun 2007 00:30:04 +0200, [EMAIL PROTECTED] wrote: On Thu, 21 Jun 2007 21:13:42 +0200, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Thu, 2007-06-21 at 10:55 +0200, [EMAIL PROTECTED] wrote: Sorry, I was not intending to emphasise layers, although I was trying to cover the case where an indexed layer was added. The basic point is that this message is great if there is an indexed element in the image, otherwise it's clutter and we could prefer to avoid displaying this for non indexed images. This whole thread could have been avoided if you had taken the time to actually try this before you write a mail. The message is only shown when scaling indexed image. If you scale an RGB or grayscale image, no such hint is displayed. Sven Very sorry to have wasted your time. Of course I did test this but I was mistaken about the nature of the image I was testing with. It was a test image in png format. I looked at image props , colour profile where it states name : sRGB , info: default RGB working space. This lead me to mistakenly think it was an RGB image. Now I double check, it is in fact an indexed image which explains the message appearing as you say. Appologies for the noise. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer Hoping to make some useful outcome to this thread ... Indexed colour layers are always scaled without interpolation. This is incorrect. I suggest the following. Indexed colour layers are always scaled using nearest neighbour interpolation. Once 2.4 is out and there is a review of the interp naming strategy w.r.t. downscaling the use of NONE should be probably be changed as well. (A scaled up image with no interpolation has holes in it.) Even though rather simplistic, N.N _is_ interpolation. ;) ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
[Gimp-developer] index colour images: interp
Hi, I just noticed a little warning has been added to interpolation dlg. Nice touch, it's important. Indexed colour layers are always scaled without interpolation. The chosen interpolation type will affect channels and masks only. I'm concerned this text is far too technical for most users and hence lost. Suggestions: c/Indexed colour layers/Indexed colour layers (eg. GIF)/ c/affect channels and masks only/only affect separate colour channels and transparency masks/ (The word reordering emphasises the only.) Indeed it may be best if this only gets shown when relevant. If there are no indexed layers present (which will often be the case) it is irrelevant and just slows the user by feeding him unneeded info to parse. Apart from that I really like the way it's layed out. The little icon brightens up the rather boring dlg. , the layout is very pleasing to the eye. /gg ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
Hi, On Thu, 2007-06-21 at 09:42 +0200, [EMAIL PROTECTED] wrote: Indeed it may be best if this only gets shown when relevant. If there are no indexed layers present (which will often be the case) it is irrelevant and just slows the user by feeding him unneeded info to parse. It is very unlikely that an indexed image doesn't contain any layers. Why do you claim that this will often be the case? Do I miss sometig obvious here? I also don't follow your other suggestions: c/affect channels and masks only/only affect separate colour channels and transparency masks/ We should probably change it to say layer masks since that's what we use throughout the user interface. But we never call them transparency masks and the term color channel is quite misleading. What's meant here is a saved selection and we call that a Channel. Sven ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Thu, 21 Jun 2007 09:58:17 +0200, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Thu, 2007-06-21 at 09:42 +0200, [EMAIL PROTECTED] wrote: Indeed it may be best if this only gets shown when relevant. If there are no indexed layers present (which will often be the case) it is irrelevant and just slows the user by feeding him unneeded info to parse. It is very unlikely that an indexed image doesn't contain any layers. Why do you claim that this will often be the case? Do I miss sometig obvious here? Sorry, I was not intending to emphasise layers, although I was trying to cover the case where an indexed layer was added. The basic point is that this message is great if there is an indexed element in the image, otherwise it's clutter and we could prefer to avoid displaying this for non indexed images. It's often the case that image is non indexed. I also don't follow your other suggestions: c/affect channels and masks only/only affect separate colour channels and transparency masks/ We should probably change it to say layer masks since that's what we use throughout the user interface. But we never call them transparency masks and the term color channel is quite misleading. What's meant here is a saved selection and we call that a Channel. Sven OK, then it's me that was interpreting those terms to mean something else. I thought this was a reference to RGB colour channels and alpha transparency. If these terms have another established meaning in the interface that's fine. layer masks would be clearer, if you think channel is unambiguous no need to change that term. corrected suggestion: c/affect channels and masks only/only affects channels and layer masks/ dont display on non indexed images. /gg ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
Hi, On Thu, 2007-06-21 at 10:55 +0200, [EMAIL PROTECTED] wrote: Sorry, I was not intending to emphasise layers, although I was trying to cover the case where an indexed layer was added. The basic point is that this message is great if there is an indexed element in the image, otherwise it's clutter and we could prefer to avoid displaying this for non indexed images. This whole thread could have been avoided if you had taken the time to actually try this before you write a mail. The message is only shown when scaling indexed image. If you scale an RGB or grayscale image, no such hint is displayed. Sven ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
Re: [Gimp-developer] index colour images: interp
On Thu, 21 Jun 2007 21:13:42 +0200, Sven Neumann [EMAIL PROTECTED] wrote: Hi, On Thu, 2007-06-21 at 10:55 +0200, [EMAIL PROTECTED] wrote: Sorry, I was not intending to emphasise layers, although I was trying to cover the case where an indexed layer was added. The basic point is that this message is great if there is an indexed element in the image, otherwise it's clutter and we could prefer to avoid displaying this for non indexed images. This whole thread could have been avoided if you had taken the time to actually try this before you write a mail. The message is only shown when scaling indexed image. If you scale an RGB or grayscale image, no such hint is displayed. Sven Very sorry to have wasted your time. Of course I did test this but I was mistaken about the nature of the image I was testing with. It was a test image in png format. I looked at image props , colour profile where it states name : sRGB , info: default RGB working space. This lead me to mistakenly think it was an RGB image. Now I double check, it is in fact an indexed image which explains the message appearing as you say. Appologies for the noise. ___ Gimp-developer mailing list Gimp-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
