DSFB and Computer Graphics: Deterministic Structural Semiotics - A Narrow Wedge for Structural Supervision of Trust-Regulated Temporal Reuse - A Deterministic Observer Layer for Graphics Diagnostics, Auditability, and Low-Risk Integration

This work introduces a deterministic supervisory layer for temporal reusepipelines, formulated as a structural residual observer operating on existing renderingoutputs. Rather than replacing reconstruction or temporal accumulation methods, theproposed Drift–Slew Fusion Bootstrap (DSFB) provides a read-only regulatory signal thatcharacterizes temporal stability, residual growth, and regime transitions at the pixel level. The system operates without modifying upstream rendering logic and is designed to benon-interfering: disabling the observer restores identical baseline behavior. A deterministicreplay contract enables exact reconstruction of supervisory signals under identical inputs, allowing pixel-level provenance and post-hoc analysis of temporal artifacts. Evaluated on real Unreal Engine captures under multiple motion regimes, DSFB demon-strates consistent supervisory behavior and, when applied in a hybrid configuration, reducesregion-of-interest error relative to a strong heuristic baseline. No claim is made of stan-dalone reconstruction performance; the contribution is a deterministic regulatory frameworkfor analyzing and augmenting existing temporal systems. This paper narrows the DSFBcomputer-graphics story to a single flagship wedge: trust-regulated temporal reuse understructural inconsistency. The verified evidence is a strict Unreal-native replay contract on five real captures from one ordered shot, using exported referencecolor as the real-reference proxy and a fixed method ladder: fixedₐlpha, strongₕeuristic, dsfbₕostₘinimum, and dsfbₚlusₛtrongₕeuristic. The headline claim is correspondingly narrow: DSFB improves strong temporal heuristics via structural supervision. The strongest current result is the fixed hybrid, which achieves Demo A ROI MAE mean ± std of 0. 00501 ± 0. 00178versus 0. 00657 ± 0. 00247 for the strong heuristic alone, while pure DSFB records 0. 04522± 0. 00683 and remains heuristic-favorable on all five captures. DSFB alone does notoutperform strong heuristic baselines in the current evaluation. The ROI definition capturesapproximately 50% of the frame under the fixed baseline-relative threshold, making themetric closer to a global structural error measure than a sparse artifact mask; the measuredROI coverage is 50. 60% ± 18. 61%. Demo B provides supportive but secondary evidence: imported trust reduces mean ROI error to 0. 23822 versus 0. 26347 for the combined heuristicand 0. 30026 for uniform allocation. Trust falls from 0. 78657 at onset to 0. 35245 at peak ROIand recovers to 0. 49284 by frame₀005, while intervention rises from 0. 21345 to 0. 64758and settles at 0. 50715, supporting trust as a useful supervisory signal rather than a universalconfidence claim. Imported-buffer GPU scaling on the measured RTX 4080 SUPER /Vulkan path ranges from 1. 0391 ms at 256×144 to 67. 7201 ms at 3840×2160; these arecompute-path measurements, not in-engine production timings. The resulting claim isnarrow, reproducible, and falsifiable rather than a broad claim that DSFB solves graphics.