Highlight Bloom and Photoillustration Look

(system) #1

Highlight Bloom and Photoillustration Look

Replicating a 'Lucisart'/Dave Hill type illustrative look

Over in the forums community member Sebastien Guyader (@sguyader) posted a neat workflow for emulating a photo-illustrative look popularized by photographers like Dave Hill where the resulting images often seem to have a sort of hyper-real feeling to them. Some of this feeling comes from a local-contrast boost and slight ‘blooming’ of the lighter tones in the image (though arguably most of the look is due to lighting and compositing of multiple elements).

To illustrate, here are a few representative samples of Dave Hill’s work that reflects this feeling:

Dave Hill Cliff Dave Hill Finishline Lotion Dave Hill Track Dave Hill Nick Saban
A collection of example images. ©Dave Hill

A video of Dave presenting on how he brought together the idea and images for the series the first image above is from:

This effect is also popularized in Photoshop® filters such as LucisArt in an effort to attain what some would (erroneously) call an “HDR” effect. Really what they likely mean is a not-so-subtle tone-mapping. In particular the exaggerated local contrasts is often what garners folks attention.

We had previously posted about a method for exaggerating fine local contrasts and details using the “Freaky Details” method described by Calvin Hollywood. This workflow provides a similar idea but different results that many might find more appealing (it’s not as gritty as the Freaky Details approach).

Sebastien produced some great looking preview images to give folks a feeling for what the process would produce:

Images from pixabay (CC0, public domain): Motorcycle, car, woman.

Replicating a “Dave Hill”/“LucasArt” effect

Sebastien’s approach relies only on having the always useful G’MIC plugin for GIMP. The general workflow is to do a high-pass frequency separation, and to apply some effects like local contrast enhancement and some smoothing on the residual low-pass layer. Then recombine the high+low pass layers to get the final result.

  1. Open the image.
  2. Duplicate the base layer.
    Rename it to “Lowpass”.
  3. With the top layer (“Lowpass”) active, open G’MIC.
  4. Use the Photocomix smoothing filter:

    Testing → Photocomix → Photocomix smoothing

    Set the Amplitude to 10. Apply.
    This is to taste, but a good startig place might be around 1% of the image dimensions (so a 2000px wide image - try using an Amplitude of 20).
  5. Change the “Lowpass” layer blend mode to Grain extract.
  6. Right-Click on the layer and choose New from visible.
    Rename this layer from “Visible“ to something more memorable like “Highpass” and set its layer mode to Grain merge.
    Turn off this layer visibility for now.
  7. Activate the “Lowpass” layer and set its layer blend mode back to Normal.
    The rest of the filters are applied to this “Lowpass” layer.
  8. Open G’MIC again.
    Apply the Simple local contrast filter:

    Details → Simple local contrast

    • Edge Sensitivity to 25
    • Iterations to 1
    • Paint effect to 50
    • Post-gamma to 1.20
  9. Open G’MIC again.
    Now apply the Graphic novel filter:

    Artistic → Graphic novel

    • check the Skip this step checkbox for Apply Local Normalization
    • Pencil size to 1
    • Pencil amplitude to 100-200
    • Pencil smoother sharpness/edge protection/smoothness
      to 0
    • Boost merging options Mixer to Soft light
    • Painter’s touch sharpness to 1.26
    • Painter’s edge protection flow to 0.37
    • Painter’s smoothness to 1.05
  10. Finally, make the “Highpass” layer visible again to bring back the fine details.

Trying It Out!

Let’s walk through the process. Sebastien got his sample images from the website https://pixabay.com, so I thought I would follow suit and find something suitable from there also. After some searching I found this neat image from Jerzy Gorecki licensed Create Commons 0/Public Domain.

The base image (link).
From pixabay, (CC0 - Public Domain): Jerzy Gorecki.

Frequency Separation

The first steps (1—7) are to create a High/Low pass frequency separation of the image. If you have a different method for obtaining the separation then feel free to use it. Sebastien uses the Photocomix smoothing filter to create his low-pass layer (other options might be Gaussian blur, bi-lateral smoothing, or even wavelets).

The basic steps to do this are to duplicate the base layer, blur it, then set the layer blend mode to Grain extract and create a new layer from visible. The new layer will be the Highpass (high-frequency) details and should have its layer blend mode set to Grain merge. The original blurred layer is the Lowpass (low-frequency) information and should have its layer blend mode set back to Normal.

So, following Sebastien’s steps, duplicate the base layer and rename the layer to “lowpass”. Then open G’MIC and apply:

Testing → Photocomix → Photocomix smoothing

with an amplitude of around 20. Change this to suit your own taste, but about 1% of the image width is a decent starting point. You’ll now have the base layer and the “lowpass” layer above it that has been smoothed:

Photocomix Smoothing
“lowpass” layer after Photocomix smoothing with Amplitude set to 20.

Setting the “lowpass” layer blend mode to Grain extract will reveal the high-frequency details:

Grain Extract HP
The high-frequency details visible after setting the blurred “lowpass” layer blend mode to Grain extract.

Now create a new layer from what is currently visible. Either right-click the “lowpass” layer and choose “New from visible” or from the menus:

Layer → New from Visible

Rename this new layer from “Visible” to “highpass” and set its layer blend mode to Grain merge. Select the “lowpass” layer and set its layer blend mode back to Normal.


The visible result should be back to what your starting image looked like. The rest of the steps for this tutorial will operate on the “lowpass” layer. You can leave the “highpass” filter visible during the rest of the steps to see what your results will look like.

Modifying the Low-Frequency Layer

These next steps will modify the underlying low-frequency image information to smooth it out and give it a bit of a contrast boost. First the “Simple local contrast” filter will separate tones and do some preliminary smoothing, while the “Graphic novel” filter will provide a nice boost to light tones along with further smoothing.

Simple Local Contrast

On the “lowpass” layer, open G’MIC and find the “Simple local contrast” filter:

Details → Simple local contrast

Change the following settings:

  • Edge Sensitivity to 25
  • Iterations to 1
  • Paint effect to 50
  • Post-gamma to 1.20

This will smooth out overall tones while simultaneously providing a nice local contrast boost. This is the step that causes small lighting details to “pop”:

Simple Local Contrast
After applying the “Simple local contrast” filter.
(Click to compare to the original image)

The contrast increase provides a nice visual punch to the image. The addition of the “Graphic novel” filter will push the overall image much closer to a feeling of a photo-illustration.

Graphic Novel

Still on the “lowpass” layer, re-open G’MIC and open the “Graphic Novel” filter:

Artistic → Graphic novel

Change the following settings:

  • check the Skip this step checkbox for Apply Local Normalization
  • Pencil size to 1
  • Pencil amplitude to 100-200
  • Pencil smoother sharpness/edge protection/smoothness
    to 0
  • Boost merging options Mixer to Soft light
  • Painter’s touch sharpness to 1.26
  • Painter’s edge protection flow to 0.37
  • Painter’s smoothness to 1.05

The intent with this filter is to further smooth the overall tones, simplify details, and to give a nice boost to the light tones of the image:

Graphic Novel
After applying the “Graphic novel” filter.
(Click to compare to the local contrast result)

The effect at 100% opacity can be a little strong. If so, simply adjust the opacity of the “lowpass” layer to taste. In some cases it would probably be desirable to mask areas you don’t want the effect applied to.

I’ve included the GIMP .xcf.bz2 file of this image while I was working on it for this article. You can download the file here (34.9MB). I did each step on a new layer so if you want to see the results of each effect step-by-step, simply turn that layer on/off:

Sample layers
Example XCF layers

Finally, a great big Thank You! to Sebastien Guyader (@sguyader) for sharing this with everyone in the community!

A G’MIC Command

Of course, this wouldn’t be complete if someone didn’t come along with the direct G’MIC commands to get a similar result! And we can thank Iain Fergusson (@Iain) for coming up with the commands:

--gimp_anisotropic_smoothing[0] 10,0.16,0.63,0.6,2.35,0.8,30,2,0,1,1,0,1

-sub[0] [1]

-simplelocalcontrast_p[1] 25,1,50,1,1,1.2,1,1,1,1,1,1
-gimp_graphic_novelfxl[1] 1,2,6,5,20,0,1,100,0,1,0,0.78,1.92,0,0,2,1,1,1,1.26,0.37,1.05
-c 0,255

This is a companion discussion topic for the original entry at https://pixls.us/articles/highlight-bloom-and-photoillustration-look/

Photoillustration look in LightZone
(Lisa Golladay) #2

Thanks for another great tutorial! One thing you might want to clarify is whether the G’MIC output mode is set to default (in place) or new layer. Your sample layer palette had me confused for a while into thinking you ran Local Contrast and Graphic Novel as new layers.

Actually, I would prefer to run them as new layers so I can play with opacity. In which case, the key step that I was missing is to make the Local Contrast layer active (and not good old Lowpass) before running Graphic Novel.

(Pat David) #3

You’re welcome - though the credit really goes to @sguyader. I just wrote up what he already described.

Ah, I thought I had mentioned it. The output is all in-place, where I will usually duplicate layers as I want. That last image of my layers was just to illustrate that I had did each step as a separate layer in the example GIMP file so that folks could fiddle with it. :wink:

(Sebastien Guyader) #4

@Acmespaceship In my workflow I set the output to “New active layer”, as I apply each new filter to the newly created layer.


I cannot download your GIMP .xcf.bz2 file . I dont know why.

(Mike Bing) #6

I see the GMIC commands here which is awesome. I like the looks of this combination of GMIC filters. Can anyone convert the commands into a bash file I could run on a whole directory of images?

(Pat David) #7

Are you downloading from the article or the forum post here?

[edit]: I see now that the link didn’t work right here in the forum. I’ve fixed it and included the link below.

In either case, here’s a direct link to the file for you! :slight_smile:


(Mike Bing) #8

Love this application of GMIC when applied to landscape images!

(Eric) #9

Thanks so much. I imagined it might work well with landscape and hadn’t had a chance to try it. Nice!

(Alex Mozheiko) #10

Wow, that’s a gem, thanks for sharing!


From the original samples it seems that the use of wide angle lenses and the distortion that comes from those help in making the images pop, too. Apart from that the look is matched quite well with those processing instructions.

(David Butcher) #12

I worked through this tutorial yesterday without any problems, and wanted to try again today.
However, when I apply the local contrast step, it results in unmistakable panels or quadrant lines showing on my photos – see attached image. What is causing this?

g/mic settings: Details->Simple local contrast
edge sensitivity 25
iterations 1
paint effect 50
channels luminance only
pre-gamma 1.00
post-gamma 1.20
all luminance masks are set to 1.00
parallel processing = auto



It is a problem with the filter not working well with parallel processing.t the turning off parallel processing or leave gamma settings at 1

(David Butcher) #14

Cheers Iain. Turning off the parallel processing cleared the problem.
Many thanks,

(Mike Bing) #15

One more for the road. This workflow really does produce interesting results!


Had the same problem with the simple contrast filter. Only after turning parallel processing off, did the tile artifact vanish.

In addition: The 1% rule for the smoothing algorithm works fine for smaller images. For larger ones (3000px+) you can bump this up. Did you make similar experiences?

(Mac Lo) #17

my test (with the direct G’MIC commands)

thanks for this tutorial.

(Mica) #18

I plan to add the gmic command (and any others I find for gmic) to our github scripts repo. :slight_smile:


Is it possible to see photos before and after?

(Pat David) #20

Are the before/after images not showing up for you? Do you mean in the article or from others comments?