2025-12-13

Scrutinizer v1.4: MIP Pooling, Oklab Saliency & Inhibition of Return

The peripheral visual system pools features over regions that grow with eccentricity. v1.4 replaces the previous 5-tap Gaussian blur with hardware MIP-map sampling, mapping eccentricity to MIP level the way receptive field size grows with distance from the fovea (validated against contrast sensitivity data; see the report). This is faster (one texture lookup vs. five) and more biologically accurate. The release also migrates saliency computation to Oklab color space and introduces Inhibition of Return as a visual memory mode.

Update (2026-10-01): Pooling-region size grows linearly with eccentricity, and crowding’s critical spacing is about 0.5× eccentricity (Bouma 1970). The v1.4 shader raised MIP level linearly with eccentricity, so its pooling size grew exponentially. The linked report tests the later DoG pipeline (v2.0–v2.1), not this mapping, and its correlation with Rovamo & Virsu (1979) fails its threshold (r = 0.60 against 0.9).

MIP-based peripheral pooling

Rosenholtz et al.’s Texture Tiling Model predicts that pooling region size grows linearly with eccentricity. Hardware MIP chains have a similar hierarchy: each level halves spatial resolution. The mapping:

// MIP level grows with eccentricity
float mipLevel = clamp(normalizedEcc * 2.5, 0.0, 4.0);
vec4 pooled = textureLod(u_texture, uv, mipLevel);

Update: This single-MIP approach was decomposed into frequency-selective DoG bands in v1.6, then upgraded to 8 half-octave bands in v2.1. The eccentricity mapping was also recalibrated to linear M-scaling in v1.8. See the MIP chain explainer for how it works now.

MIP pooling is ~5x faster than the Gaussian blur it replaces, with smoother transitions and better color fidelity in the periphery. Later releases built on this MIP pipeline.

Oklab saliency

Saliency feature extraction moved from RGB to Oklab, separating lightness (L) from chrominance (a, b) in a perceptually uniform space. This mirrors the magnocellular (luminance) and parvocellular (color) pathways. In practice, saliency maps are more stable and less sensitive to hue-saturation confounds.

Update (2026-10-01): Oklab is a perceptual color space. It does not model the magnocellular or parvocellular pathways.

Inhibition of Return

A new visual memory mode: recently fixated areas become suppressed (more distorted), simulating the brain’s tendency to de-prioritize previously attended locations. This pushes gaze toward novel content and complements Scrutinizer’s existing memory-based clearing.

IOR is one of the mechanisms behind efficient information foraging (Pirolli & Card 1999). By suppressing already-visited locations, the visual system biases scanning toward unexplored regions, favoring exploration over exploitation.

Update (2026-10-01): Pirolli & Card’s information foraging theory does not address inhibition of return. The proposal that IOR aids visual search is Klein & MacInnes (1999), and it is disputed for saccadic search and free viewing (Hooge et al. 2005; Smith & Henderson 2009).

Other changes

Visual Overlay refinement: Renamed from "debug grid," switched to linear spacing with variable stroke width.
Coupled Warp + MIP Pooling (v1.4.1): Position jitter scales with integration field size for physically-grounded crowding simulation.
Unbound Color (v1.4.1): Parvocellular resolution loss modeled as chromatic fringe blur ("watercolor bleed") with radial offset.
Auto-updates: Checks GitHub releases on startup.

Links: GitHub · Full changelog · v1.4.0 release