E.g. try Remove Grain...
Is this video or a still? The whole image is full of huge compression artifacts. Even the clouds are blocky. Softening or adding noise sufficient to remove these artifacts is going to seriously degrade the image. It if's a still try finding another one. If it's video the first think I'd do would be to transcode the video to a 10 bit lossless codec and then start applying filters to the transcoded file.
It's time-lapse video.
It's going to be used in a heavily stylised sequence so medium-ish quality is acceptable, but it would be nice to clean the image up before proceeding.
Noise removal helps, but it's a bodge. I'm tempted to key out and replace the sky.
Why would you transcode first rather than just feed it straight into AE?
Any chance this is a raw sequence? Sometimes I've had luck with Selective Color and/or Hue and Saturation controls in PS, in the ACR/AE interface you might find relief in HSL/Grayscale and Noise reduction for the chromatic aberration issues. Unfortunately the blocky artifacts are beyond my expertise...
Good luck, the light is quite nice in this still!
Any time you're working with compressed sources it's best to transcode before doing any significant processing. If you're on a Mac I'd suggest Apple PRO Rez 422, if on a PC try one of the Black Magic 10 bit codecs.
Does this image look this compressed in Photoshop???? Is it part of an image sequence? If the image sequence looks better in Photoshop or Lightroom you may be able to batch process the image sequence to a better format.
Probable off topic...
Rick, could you point me to any theory why transcoding is better than working with a native source?
Interesting. I'll try that. Thanks.
Unfortunately the source material is long gone.
If you're working with an 8 bit file you only have 256 levels of color for each chanel. When you try Let's say that you're trying to smooth out a change in the blue that goes from 200 to 210 over 300 pixels. That means that you're going to have 30 steps in the blue in those 300 pixels and they are going to be visible.
If your source has been transcoded to 10 bit now you have 1024 color values available, the change in color over those 300 pixels will be calclcuated in 120 steps or about every 2 pixels. The original still has a shift in value every 30 pixels but there's now more room to smooth out the differences.
There's more to it than just the extended color space. Highly compressed video streams make guess at the color data for the frames that are inbetween the i frames. At a minimum, 1/3 of all frames are compilations of the data from the only real frames recorded. All 10 bit codecs used in production use every frame. When you transcode you turn that fabricated data into real frames. This gives you better data to work with especially when you're manipulating the color chanels. It's just better all the way around.
Transcoding alone won't do anything to the pixel values, it just gives any compositing or re-compressing app a better chance to perform the magic. I transcode everything that I'm going to heavily process including XDCam and Panasonic P2 footage. DSLR footage is also always transcoded using Magic Bullet Grinder.
Thanks Rick, I got your point of view.
Actually, I noticed that in some posts earlier.
It's probably interesting how it's differ to e.g. Colin Brougham's one since Adobe CS applications are able to process compressed footages natively...
Ah. Thanks. I'll look into that.