Forums >
Digital Art and Retouching >
HighPass Sucks (+ solution)
I read about 15 pages of this and my head just hurts. Not sure I even learned anything. I think I'm confused since the technique spread out into different categories like sharpening and skin smoothing. And I would regard myself as very good with photoshop lol. So I perhaps am just confused on what is being done to each layer to sharpen, or smooth skin or what have you. Perhaps I'm confused because I'm a visual learner and would prefer to see someone doing it in a youtube video. Anyone wanna do that! The intro post explains how to lay it out, but to me is a bit blurry when explaining what what to do to which layers after you have got your low-frequency and high-frequency layers. Further it seems now if someone is confused then the people who are not so confused are just telling us to go back and read again. I apologize if it is obvious or makes sense, but I would love to see this layed out a bit more simple and dumbed down once again even though it has been, or a video made to show exactly how to sharpen or how to smooth skin using the technique. I seriously felt like I was switching back and forth between what people were trying to do with the technique every page or so haha. Anyhow, I'm very interested and curious. And I've been reading for three hours. But have not figured anything out! Aug 11 10 07:40 am Link I have gone back and re read, and get it slightly more. But would still love to see a video. A video I would even pay some monies for. Aug 11 10 08:00 am Link Just the page before this one gives you two videos: http://www.model-citizens.com/Tutorials … ion-1.flv, http://www.youtube.com/watch?v=kcY0Anrd5Yk (maybe this one as well: http://www.youtube.com/watch?v=2D2GZgUs9BQ although not the best example) and here's another one: http://www.digitalphotoshopretouching.c … ir-retouch If you've still got some questions left, feel free to ask :-). Best regards, Jonas Aug 11 10 08:09 am Link yea those helped quite a bit thanks. Guess I should of went all the way instead of stopping at like 15 haha. Just seemed like it was getting more complicated when I didn't even understand the basics. Aug 11 10 08:29 am Link I Must Be Dead wrote: 3 hours? Try 14 months. Aug 11 10 09:06 am Link DerW wrote: The first link has a added comma in the link http://www.model-citizens.com/Tutorials … tion-1.flv Chain;271603 wrote: Aug 12 10 06:21 am Link SRB Photo wrote: So I get the advantage of splitting the image in two layers but what's the advantage of splitting it in three? A more targeted band pass filter, sharpening? Aug 24 10 09:31 pm Link Ruben Vasquez wrote: Pixel-for-pixel reconstruction of the original image. For most people and most purposes it wouldn't be a meaningful difference vs. splitting to two layers, but for the forensic community and the truly anal it might make a difference. For the rest, it's just an interesting bit of understanding about how PS works to know. Aug 25 10 02:09 am Link Great examples of the power of the FS: FLEXmanta wrote: Aug 25 10 04:49 am Link SRB Photo wrote: Ok. So for general frequency seperation type moves, two layers is fine then? Aug 25 10 07:45 pm Link Ruben Vasquez wrote: Of course. Split, work, merge... split again later, work some more, merge again. You can choose a different radius every time and as many times you want. That's why the high pass sucked, cause you coudn't, and now you can because the split is identical to the merged. The split is one of those techniques that works well when you work on a "pixels on the bottom, adjustments on top" basis. Aug 26 10 09:14 am Link Just finished reading this thread again man so many techniques that are shown and discussed has me writing them down in my ps journal of variations to test and try out that I missed or skimped over not understanding the benefits but see it a bit more clearly now. Funny when the light clicks on all of sudden you see possibilities where before you saw none... Aug 26 10 02:03 pm Link Ruben Vasquez wrote: For 99.9% of work anyone is going to do, absolutely. I don't believe in doing 2 layers for 8bpc editing, but I accept that some do and still get results which are acceptable for them. OTOH, for those working in 16bpc, using >2 layers for a single-frequency split would, IMO, be over the top. Aug 26 10 02:06 pm Link Is there a way to do a frequency seperation in LAB? I don't seem to be having much luck when I attempt it. Sep 06 10 10:08 pm Link Ruben Vasquez wrote: It can be done - it's just a PITA for the way PS handles LAB blend modes and its Brightness / Contrast implementation. Sep 07 10 02:52 am Link Hi @all. I have been using this method for a while now and am really happy with the results. I do have two questions though and am hoping somebody can help Sometimes when editing the high layer close to edges (for example removing stray blond hairs close to the hairline in shadow areas) I get color smears. How can this be explained and what am I doing wrong? I am using a hard edged brush and my understanding is that color smears shouldn't appear anyway because there is no color information on the high level. I've went through it over and over again, following the instructions here carefully. Anybody have an idea what is going wrong? Second question: what influence does the chosen radius for gaussian blur (low level) have on apply image (high level)? Any help appreciated. Thanks in advance. Greetings from Germany Andy Sep 09 10 05:38 am Link picturepipeline wrote: I'm guessing you're talking about using the healing brush near edges. Sep 09 10 06:40 pm Link To build on what P2P said, also make sure that your Heal / Clone tool is set to sample the Current Layer Only. Also, he is also correct - all RGB color images will have color in the HF data - it's just a part of the tools we're using. Sep 09 10 06:44 pm Link Pixels 2 Pixels wrote: Good points, mentioned above. Sep 09 10 06:45 pm Link Thanks for the replies to my post. As mentioned I am sticking exactly to the instructions, using hard edged brushes and working on the current layer only. The methods mentioned are exactly the methods I use to deal with the problem. The smears actually aren't a real problem - just annoying and I therefore wanted to know if I had overlooked something and/or am doing something wrong. Watching different videos/tutorials and Natashia Taffarels DVD on the topic I had won the impression that these smears should not occur in the first place, because color and texture are seperated on different layers (HF/LF). So it appears that that was my misunderstanding. The smears are the "normal" behaviour I would expect from the healing brush and I thought this problem was avoided by the split. I usually don't use the healing brush on the HF layer. I don't feel that the sampled (source) texture needs to be fit in by calculations. The clone tool with a hard edged brush (95-100%) works well for me. Thanks again for the replies. You all have a great day and a great weekend! cu around Andy Sep 09 10 11:13 pm Link Krunoslav-Stifter wrote: I've tried the selection thing and it still pulled information from the edge (outside the selection). I'll have to look into that again. Sep 10 10 03:00 am Link Pixels 2 Pixels wrote: It's a tip that you can actually find in Photoshop official help manual. Sep 10 10 03:19 am Link Pixels 2 Pixels wrote: I have experienced the same bug using a selection maybe ithe selection was feathered don't know always switched to cloning or changed heal brush to replace but a pain since I forget to switch it back and wonder why it's not working right lol.... Sep 10 10 05:11 am Link SRB Photo wrote: Thanks Sean! Iv been gone for awhile so I'll have to play with this for sure. I'm a little suprised by the methods though. I would have thought the technique would focus on the L channel only. Do you still have the advantage that LAB offers, that of seperating color and contrast? Sep 19 10 07:41 am Link Ruben Vasquez wrote: You can do it that way as well - applying the 50% gray fill to the L* channel alone and fading. Sep 19 10 12:10 pm Link SRB Photo wrote: Any particular reason why? Sep 19 10 04:06 pm Link Ruben Vasquez wrote: The color and luminosity / lightness interact in life to create images which look natural to us. The same is true when you do your frequency separations. There will be hue shifts during the transition from highlight to shadow, saturation differences between these and between midtones, etc. By completely separating color from luminosity, I find that the result often has color inconsistencies on close inspection - places where those natural variations have become inconsistent with their corresponding luminous pixels. Certainly there may be situations where it's advantageous to do so - I've just never found it to be so when working with people. Sep 20 10 03:37 am Link WOW it's been LONG time coming long live Sean the man :-).lol Sep 20 10 11:54 am Link SRB Photo wrote: I can think of a couple. Typically lips are much more positve in the A channel then the rest of the skin which allows some wiggle room for ajusting saturation of just the skin while maintaining saturation or adjusting to taste, the saturation of the lips. Also dark circles under the eyes are typically less saturated then the rest of the skin which is easy to adjust in LAB without affecting the rest of the face or the need of masks. Sep 20 10 09:07 pm Link I have learned more from this thread than I could ever hoped, seriously, thank you to everyone involved The information is vital in here, just wow Oct 03 10 12:08 pm Link so...not gunna lie..feel like someone punched the daylights outa me..head is spinning. Some of you all are sooo amazing at what you do. This thread is AMAZING. hopefully in a few more weeks of re-reading this thread ill understand it ... Oct 25 10 09:43 pm Link I'm just reading the thread over at RetouchPRO and thought I might throw in an idea or two of mine as well :-). The apply image technique is nothing I would change at all, it works great! The B/C idea however I'm still not sure about (I'll come to that in a minute). Instead of using the B/C (applied directly) to split the frequencies (or for that matter create a low pass and a band stop in the next step), I'm thinking about using 50% fill opacity instead. Using the fill has the benefit that some values aren't clipped that much (as you might know, fill calculates the blending with the correct data, instead of opacity, which uses clipped data), so in my tests (I checked with the histogram) the fill was a little more accurate than the B/C (very little though). Maybe it's worth checking and I'd be very interested to see what you think about it :-). Best regards, Jonas Nov 02 10 08:49 am Link DerW wrote: Jonas - how are you applying the 50% fill to create the effect? Inverted, blurred copy overtop in LL with 50%F, or something else? Thanks. Nov 02 10 09:38 am Link Exactly. I uploaded you a PSD-file, that should make the whole thing a little clearer: http://www.sendspace.com/file/ewmcn9 With the fill opacity the difference to the low pass is mean: 0.33, Std. Dev: 2.94; with B/C the mean is 0.34, Std. Dev: 3.03 (fwiw, the difference to the low pass with normal opacity is mean: 1.1, Std. Dev: 4.67). The test image was created in LAB with gradients from black to white in the a- and b-channels and a radial gradient in the L-channel. I switched to RGB after that, set the saturation to +100 and added difference clouds as well as (color) noise. I tested this on a few more images however, but always got the same results and thought that this kind of synthetic image could be a little more telling ;-). Best regards, Jonas Edit: As I said, I'm not sure about this, so I thought you might want to test it as well. Didn't know you already did ;-) Nov 02 10 10:36 am Link DerW wrote: Can you tell me how you generated your B/C layer? I've looked at it and agree that the version you're showing has a considerable difference from 'right' due to clipping, but when I create the B/C version myself get replication within 1/32k points in all channels. Let me know! Nov 02 10 11:01 am Link I duplicated my background layer, inverted it, set the blending mode to "Linear Light" and used "Image"-"Adjustments"-"Brightness/Contrast" with contrast set to -50 and use legacy turned on (first I feared I might have used the adj. layer, but I replicated it with direct use of the B/C and it turned out to be the same). Now I used the "Filter"-"Other"-"High pass" on it with a radius of 20Px. All in 16Bit. Though the difference is extremely small, I'm still wondering why you get different results from mine? Best regards, Jonas Edit: Okay, now I'm really confused. I redid it one last time and now the results were perfect. Maybe it has to do with the order in which the B/C is applied? Edit #2: Yes. The B/C has to be applied before the high pass, otherwise results will differ a little. Okay, problem solved, forget the fill ;-). Nov 02 10 11:37 am Link lol. You solved it while I was telling you the same thing . Love the dedication! Nov 02 10 11:44 am Link Hehe, I wasn't reading the RP-thread while I was testing here, so sorry about that ;-) Best regards, Jonas Nov 02 10 11:48 am Link DerW wrote: How could I blame you for not reading that thread? My head hurts trying to deal with the fact inoculation which seems to have gone on in there. Nov 02 10 11:51 am Link And as a note for the MM crowd, sparked by an incessant conversation elsewhere, it occurred to me that yet another way of halving image contrast prior to employing the native High Pass filter is to use the Custom filter with a matrix of: 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 Scale 2 Offset 0 The advantage is that this is an extremely efficient way of doing things (or at least, should be - I won't go out on a limb and vouch that a filter this old is actually optimized) and is another 100% Smart Object-compatible solution. Nov 02 10 07:40 pm Link |