The paper presents a computer vision system that can recognize three-dimensional objects from unknown viewpoints using single grayscale images. Unlike other approaches that attempt to reconstruct depth information, this system uses three main mechanisms: perceptual organization, probabilistic ranking, and spatial correspondence. Perceptual organization forms groupings and structures in the image that are invariant across different viewpoints. Probabilistic ranking reduces the search space by ranking potential matches based on their reliability. Spatial correspondence solves for the unknown viewpoint and model parameters by matching projections of 3D models to the image. The system, named SCERPO, demonstrates robustness in the presence of occlusion and missing data through the application of viewpoint consistency constraints. The paper also discusses the role of depth reconstruction in human vision and argues that it is not a primary pathway for recognition. Instead, the system relies on perceptual organization and probabilistic reasoning to achieve robust recognition. The final judgment of object presence is based on the degree of overconstraining the viewpoint estimate, with correct matches typically having over 20 image segments in close agreement with the model.The paper presents a computer vision system that can recognize three-dimensional objects from unknown viewpoints using single grayscale images. Unlike other approaches that attempt to reconstruct depth information, this system uses three main mechanisms: perceptual organization, probabilistic ranking, and spatial correspondence. Perceptual organization forms groupings and structures in the image that are invariant across different viewpoints. Probabilistic ranking reduces the search space by ranking potential matches based on their reliability. Spatial correspondence solves for the unknown viewpoint and model parameters by matching projections of 3D models to the image. The system, named SCERPO, demonstrates robustness in the presence of occlusion and missing data through the application of viewpoint consistency constraints. The paper also discusses the role of depth reconstruction in human vision and argues that it is not a primary pathway for recognition. Instead, the system relies on perceptual organization and probabilistic reasoning to achieve robust recognition. The final judgment of object presence is based on the degree of overconstraining the viewpoint estimate, with correct matches typically having over 20 image segments in close agreement with the model.