In this video, we're looking at noise reduction in DaVinci Resolve. This feature requires a studio license. First, a quick primer. All video images contain noise. A certain amount looks natural, but excessive noise can detract from an image and make grading operations difficult. Most commonly, you'll encounter noise when working with underexposed or dark footage. That's because the part of the image that you want to keep, the signal, is so dark that it gets mixed in with the noise. Resolve's noise reduction tools are found on the color page and must be applied in a node. It's best practice to apply your noise reduction in the first node at the beginning of your node tree. Why? Because it's designed to work with your camera image. Applying noise reduction after color grading may have an adverse effect on the noise reduction operation. Additionally, many color grading operations benefit from a cleaner, noiseless image. Please note, if you do decide to add noise reduction to the beginning of your node tree after you've started working on your image, that's fine as noise problems aren't always obvious until you've made your primary adjustments. But if you do, just check if you need to finish downstream adjustments. There are two types of noise reduction in Resolve. Temporal and spatial. Let's start with temporal. By examining the difference between consecutive frames, temporal noise reduction attempts to work out which parts of the image are noise and which parts are signal. The easiest way to understand temporal noise reduction is to imagine a completely locked off shot with no movement in the frame. In this scenario, the only part of the image that will change is the noise. So, by process of illumination, noise can be identified and removed. This process, however, is complicated due to the fact that most video images contain movement. So Resolve uses motion estimation to reconcile the movement between frames before identifying noise. Spatial noise reduction works differently. It looks within the frame, examining pixels in a defined radius, trying to identify what is noise and what is not. Without the context of examining adjacent frames, it's easier for spatial noise reduction to incorrectly identify wanted textures in your image as noise. As a result, it's easier to accidentally soften your image with spatial noise reduction. To apply temporal noise reduction, select the number of adjacent frames you would like it to analyze. Higher values are good for shots with less movement. Lower values are recommended for shots with more movement. Changing motion estimation type to better will increase render times but improve the quality of the noise reduction. But in this instance, as the shot is completely locked off, it's not needed, so I'll get an improved result if I disable motion estimation. Finally, activate the noise reduction by changing the Luma and Chroma threshold. By default, these two controls are linked. Similarly, chroma noise is less attractive than luma noise. Unlinking these controls and applying more chroma noise reduction than luma will often result in a more pleasing image. To apply spatial noise reduction, again increase the luma or chroma threshold. As you can see, aggressive values have an adverse effect on the image's sharpness. Mode and radius can be fine-tuned to suit the needs of your image. In this instance, enhanced preserves more detail. It will affect render time though. In some instances, you might choose to use temporal and spatial noise reduction simultaneously. Let's check out one more image. A combination of both might be the best solution to address your noise issues. Noise reduction is computationally intensive, so it's pretty common to not get real-time playback. An easy fix for this is to set render caching to smart. Nodes with noise reduction are automatically marked red for caching. Once cached, it will turn blue, and once again, you'll get smooth playback. Balance is needed. Noise reduction should be applied so as to remove unwanted noise, but not degrade your image. In most cases, your aim is not to completely remove noise, but to lower the noise of problematic shots to a level that matches the levels of noise present in the rest of the sequence. You might need to decide what looks best, removing details or leaving some noise. I hope you found this video helpful. If you did, give it a like and subscribe for more.
