RECURSIVE VISION COMPUTING: A SELF-REFERENTIAL FRAMEWORK FOR ADAPTIVE PERCEPTION

Recent advances in vision computing have enabled artificial systems to achieve strong performance in image recognition and visual understanding through deep learning. However, contemporary architectures primarily operate as statistical pattern processors and lack persistent internal models of the observing system, limiting adaptability and contextual reasoning in dynamic environments. This paper proposes a Recursive Vision Computing framework that introduces a selfreferential architectural layer allowing vision systems to model their own perceptual processes alongside external visual inputs. Integrating principles from information theory, recursive systems, and machine learning, the framework establishes an observer-aware computational structure in which perception and self-evaluation evolve through continuous feedback. Rather than treating vision as a simple input–output transformation, perception is formulated as an adaptive and iterative interaction between environmental information and internal system representation. A conceptual architecture, algorithmic workflow, and experimental design are presented to demonstrate how recursive feedback may enhance adaptability, interpretability, and robustness in vision-based AI systems. The proposed approach provides a foundation for next-generation vision computing systems supporting reflective learning and context-sensitive perception, with applications in robotics, autonomous systems, and human–computer interaction.