Abstract:Handling imperfect information problems is fundamental to perception, learning, decision-making, and other processes in the environment with omnipresent incompleteness. Ensemble perception, or the ability to extract statistical properties of stimulus clusters at a glance to condense information, has been proposed as a way to partially overcome the imperfect information problem by providing global clues. However, ensemble perception itself suffers from the imperfect information problem: Due to capacity and environmental constraints, not all elements in a cluster are readily accessible, resulting in incomplete data that is required for computing statistical features. In this case, the internal correlations of those clusters could serve as an effective clue (i.e., regularity) to shed light on the missing pieces. To test this hypothesis, we investigated the role of spatial regularity in ensemble perception under imperfect information situations. We created imperfect information scenarios by using partially occluded stimulus clusters. In two experiments, we manipulated stimulus size in a cluster to linearly vary with its location. Therefore, the spatial regularity supplied clues for inferring the statistics of the parts that were not visible. The participants were instructed to estimate the average size of the entire cluster, including both the visible and invisible parts hidden behind an occlusion. We observed robust biases toward the overall cluster in the participants’ estimation, implying that the invisible parts were considered during ensemble perception. We proposed that this effect could be understood as a result of Bayesian inference, assessing evidence from the visible parts to construct the missing parts required for ensemble perception. We built a generative model, the Regularity Based Model, to simulate the Bayesian inference, which performed better in capturing the pattern of human outcomes than the comparative models. These findings indicate that the visual system could use high-level structural information to infer scenes with incomplete information, thus, producing more accurate ensemble representations.

Structure-guided Image Completion Via Regularity Statistics

Structure-Aware Image Completion with Texture Propagation

Structure-Guided Image Completion With Image-Level and Object-Level Semantic Discriminators

A Non-Local Method For Robust Noisy Image Completion

Efficient Image Completion Method Based on Alternating Direction Theory.

Structure Preserving Large Imagery Reconstruction

Automatic and accurate image completion from a large displacement view

Structure-Guided Image Inpainting Using Homography Transformation

Completeness out of Incompleteness: Inferences from Regularities in Imperfect Information Ensembles.

Low-Light Image Enhancement Via Structure Modeling and Guidance

A Structure-Preserving Image Restoration Method with High-Level Ensemble Constraints

Boosting Structure Consistency for Multispectral and Multimodal Image Registration

Point cloud completion network for 3D shapes with morphologically diverse structures

Amodal Layout Completion in Complex Outdoor Scenes.

Completeness Out of Incompleteness Inferences and Employment Regarding Regularities in Imperfect Information Ensembles

Structure-Aware Image Expansion with Global Attention

Image Tag Completion by Low-Rank Factorization with Dual Reconstruction Structure Preserved

Image completion using structural priority belief propagation.

Low-rank image tag completion with dual reconstruction structure preserved.

SRPCN: Structure Retrieval based Point Completion Network

Im2Struct: Recovering 3D Shape Structure from a Single RGB Image