cpc
Well-known member
Hey Kyron, thanks for supporting slimRAW.
The amount of information culled by the lossy compression modes depends a lot on the amount of information in the original image. Note that information is not the same as data: the purpose of compression is to more effectively represent information into less data. For example, for the same resolution, the image of the Blackmagic cameras will have higher entropy than the image of the FS cameras. So more information is discarded when doing lossy compression on BM raw images compared to FS raw images. On the other hand, BM cameras exhibit significantly more noise and aliasing which helps mask image changes caused by lossy compression.
One way to think about this is the following: the better a raw image compresses when using lossless compression, the less information needs to be discarded when doing lossy compression on the same image. To illustrate the case above, BM cameras produce images that usually compress losslessly to around 1.5:1-1.6:1. Sony FS cameras produce raw images which usually compress losslessly to better than 2:1 (sometimes up to 2.8:1). Hence, less information is discarded when doing lossy compression on Sony FS images.
Since the lossy compression in slimRAW is fairly light and perceptually based (what's commonly -- and somewhat arbitrarily -- touted by marketing as "visually lossless"), ideally degradation will be negligible. I can say I am pleased with how it turned out. You should run tests though.
Compared to ProRes422HQ, 4:1 will produce slightly smaller files (10-12% smaller). The main advantage of using raw compression vs ProRes is that you can still do a full raw workflow (control debayer quality, adjust WB, etc). Also, ProRes422HQ is 10-bit and CinemaDNG compression is 12-bit. And ProRes operates on full RGB samples (vs Bayer samples in raw compression), which puts it in a bit of a disadvantage sizewise. On the other hand, lossy CinemaDNG is only supported by Davinci Resolve at the moment.
The amount of information culled by the lossy compression modes depends a lot on the amount of information in the original image. Note that information is not the same as data: the purpose of compression is to more effectively represent information into less data. For example, for the same resolution, the image of the Blackmagic cameras will have higher entropy than the image of the FS cameras. So more information is discarded when doing lossy compression on BM raw images compared to FS raw images. On the other hand, BM cameras exhibit significantly more noise and aliasing which helps mask image changes caused by lossy compression.
One way to think about this is the following: the better a raw image compresses when using lossless compression, the less information needs to be discarded when doing lossy compression on the same image. To illustrate the case above, BM cameras produce images that usually compress losslessly to around 1.5:1-1.6:1. Sony FS cameras produce raw images which usually compress losslessly to better than 2:1 (sometimes up to 2.8:1). Hence, less information is discarded when doing lossy compression on Sony FS images.
Since the lossy compression in slimRAW is fairly light and perceptually based (what's commonly -- and somewhat arbitrarily -- touted by marketing as "visually lossless"), ideally degradation will be negligible. I can say I am pleased with how it turned out. You should run tests though.
Compared to ProRes422HQ, 4:1 will produce slightly smaller files (10-12% smaller). The main advantage of using raw compression vs ProRes is that you can still do a full raw workflow (control debayer quality, adjust WB, etc). Also, ProRes422HQ is 10-bit and CinemaDNG compression is 12-bit. And ProRes operates on full RGB samples (vs Bayer samples in raw compression), which puts it in a bit of a disadvantage sizewise. On the other hand, lossy CinemaDNG is only supported by Davinci Resolve at the moment.
Last edited:
