Odilon addresses the "text degradation" artifact inherent in traditional full-screen accessibility filters. By implementing a Semantic Segregation architecture, the engine parses the active Document Object Model (DOM) to isolate raster and vector graphical elements. This allows for the targeted application of color correction solely to media nodes (img, video, canvas, svg), strictly preserving the luminance contrast of typography.

The correction pipeline utilizes a modified Machado et al. (2009) algorithm operating within the linear-light LMS color space. By modeling human cone spectral sensitivity, the system calculates the error vector between the simulated dichromatic perception and the original signal. A compensation matrix is then computed and projected back onto the sRGB gamut to spectrally shift confusing hues into distinguishable bands without altering luminance.

Built on a modern Manifest V3 architecture, Odilon injects GPU-accelerated SVG matrix filters (feColorMatrix) directly into the browser compositor. This approach ensures zero-latency processing of dynamic content, including HTML5 video and WebGL contexts, without inducing layout thrashing or blocking the main JavaScript thread.