vsTTempSmooth is a high-performance, motion-adaptive temporal smoothing filter designed for VapourSynth and Avisynth+. It is a modernized fork of the classic TTempSmooth plugin, rewritten specifically to implement native high bit depth processing (8-16 bit integer and 32-bit float) and support advanced color spaces. Core Mechanics & Features
Unlike standard temporal blurs that blindly smear frames together, vsTTempSmooth selectively targets noise based on localized motion:
Motion Adaptive Architecture: It performs denoising exclusively on the stationary or slow-moving parts of an image, preventing ghosting, trailing, or blending artifacts on fast-moving objects.
Temporal Radius Calibration: Users can set a maximum temporal radius (maxr) from 1 to 7. At maxr=7, the filter analyzes up to 15 total frames (7 forward, 7 backward, plus the current frame) to calculate a pixel’s weighted average, vastly improving image stability.
Inverse Distance Weighting: The filter applies pixel weight based on temporal distance and value deviations. Pixels slightly under the threshold value carry less weight in the average, which mitigates sharp artifacts.
Pre-filtered Clip Support (pfclip): Users can pass an external, heavily filtered clip (e.g., heavily blurred or motion-compensated) to calculate the motion thresholds and scene changes while applying the actual smoothing directly to the clean, original source. The Evolution: Why Native High Bit Depth Matters
The original Avisynth TTempSmooth was heavily limited to standard 8-bit YUV processing. Implementing native high bit depth (10-bit, 16-bit, and 32-bit float) via vsTTempSmooth solves major fidelity issues in video encoding: View source for VsTTempSmooth – Avisynth wiki
Leave a Reply