Posts, page 11

Animation Compression: Error Compensation

Unless you use the curves directly authored by the animator for your animation clip, the animation compression you use will end up being lossy in nature and some amount of fidelity will be lost (possibly visible to the naked eye or not). To combat the loss of accuracy, three error compensation techniques emerged over the years to help push the memory footprint further down while keeping the visual results acceptable: in-place correction, additive correction, and inverse kinematic correction.

Read the full post

Animation Compression: Signal Processing

Signal processing algorithm variants come in many forms but the most common and popular approach is to use Wavelets. Having utilized this method for all character animation clips on all supported platforms for Thief (2014) I have a fair amount to share with you.

Read the full post

Animation Compression: Curve Fitting

Curve fitting builds on what we last saw with linear key reduction. With it, we saw that we leveraged linear interpolation to remove keys that could easily be predicted. Curve fitting archives the same feat by using a different interpolation method: a spline function.

Read the full post

Animation Compression: Linear Key Reduction

With simple key quantization, if we needed to sample a certain time T for which we did not have a key (e.g. in between two existing keys), we linearly interpolated between the two.

Read the full post

Animation Compression: Sub-sampling

Once upon a time, sub-sampling was a very common compression technique but it is now mostly relegated to history books (and my boring blog!).

Read the full post