All Things Color for Film and Digital Cinema
Subscribe | Log in

Posts Tagged ‘Red’

“Needle” – John Soto



Director - John Soto

DOP - Steve Windon ACS

Grading Platform - Baselight 4

Format - RED

Before Grade

After Grade

Before Grade

After Grade

Before Grade

Before Grade

After Grade

D-Lux. Website Live


See Website for more details.

Adding Grain to Digital images – increased perceptual Sharpness

Adding Grain to Digital images to enhance Perceptual sharpness and possibly perceptual color separation.

Here are some new stills from the ‘Protected’ post that was originally pulled due to image content problems. I am in the process of refactoring the original post to discuss MTF and related ‘sharpness’ issues.

Below I have posted some crappy low res, fast loading stills with links to the full res images. You most likely wont see a visual difference in the imagery looking at these jpegs.

First off is a full frame image showing the cutout/ image extraction that i have extracted for the following examples.


For reference I have also included a white guideline around the original images within which the grain has been overlayed. Outside the lines remain unaffected.

himnograin himgrain

Links – Man No Grain     Man With Grain     (The Links are quite big files in PNG format.)

All the images bar one are from a RED shoot. As a reference I have also inserted a 35mm super 2k scanned image. I have on purpose included a ‘head shot’ of a similar photographic size and taken the exact same extraction from the original frame.


Links – 35mm Reference

I felt while grading the RED job that I constantly wanted to add a textural content to the images, because in essence from my POV they were A: visually soft in comparison to film and B: way too clean, in a videoish sense.

hernograin hergrain

Links – Woman No Grain           Woman With Grain

By finely adding a moving granular overlay to the images I believe the result looks sharper ‘perceptually sharper’ that the original. The result I found was better than just sharpening the overall image which, although helps in bringing out skin tone detail also potentially causes aliasing problems later and can if used incorrectly make the image jut look like sharp video.

The addition of grain also has the effect of dithering color and greyscale content.

The idea of adding grain to video images is nothing new and has been done for years as an effect in making video and CGI imagery look more filmic or integrate more seamlessly within a predominantly film based project. This though I believe is a different take on the method where we are using it to create a visual sharpening of the images and hence add depth to otherwise ‘flat’ imagery.


Resize methods – shooting Red 3K for Post.

Over the weekend the were a few posts on the CML forums regarding shooting Red with non standard 3K files for post.  (Film standard being 2k or 4k).

My theory was that 3k is a non standard production/deliverable format and hence would at some stage have to be resized to fit standard deliverable Specs such as 2k or 4k film or HD for Video. In terms of downressing a 3k image digital filters must be applied to the pixels so an interpretation can be made of surrounding pixels to determine what pixel value is written. 

With this test I have taken a 2k reference image (I dont have any 3k reference files at hand) and performed a similar 1.5 scale reduction (emulating 3k to 2k) using different resize algorythms to see what the result would be like. I have chosen standard ‘out of the box’ filters available in most DI tools. Most of these filters inherently end up with artifacts that must be tested on the pictures at hand to see what filter best suits the production. The artifacts to look out for are Aliasing, blurring and Edge ‘Halo’ and jagged diagonals.

Unfortunately I dont have a Sinc filter at hand that is considered to be the best Downressing filter as it keeps small details without the ‘ringing’. I will add this to the data as I get a chance to go into work.

See Below. Click on each thumbnail to see the 100% cropped rendering of the transformation.

Here is a composite image of the comparison Data. 

Of particular note is the lack of all methods to efficiently interpret any of the ’40 lines’. In all cases this data, as ‘resolution’ is lost. You can also see interesting effects at each of the other resolution interpretations.  


3k to 2k resize comparison 

3k to 2k resize comparison

The basic conclusion looks like youd be better off shooting 2k but I will wait until I try the Sinc Filter. Upressing methods would probably give better results but then why not shoot 4k? 

Adrian Hauser