文档介绍：Context Based Object Categorization: A Critical Survey Carolina Galleguillos 1Serge Belongie 1,2 1Computer Science and Engineering, University of California, San Diego 2Electrical Engineering, California Institute of Technology {cgallegu,sjb}***@ goal of object categorization is to locate and identify in- stances of an object category within an image. Recognizing an object in an image is di?cult when images present occlusion, poor quality, noise or background clutter, and this task es even more challenging when many objects are present in the same scene. Several models for object categorization use appearance and context information from objects to improve recognition accuracy. Appearance information, based on visual cues, can essfully identify object classes up to a certain extent. Con- text information, based on the interaction among objects inthe scene or on global scene statistics, can help essfully disambiguate appearance inputs in recognition tasks. In this work we review di?erentapproaches of using contextual information in the ?eld of object categorization and discuss scalability, optimizations and possible future approaches. 1 Introduction Traditional approaches to object categorization use appearance features as the main source of information for recognizing object classes in real world images. Appearance features, such as color, edge responses, texture and shape cues, can capture variability in objects classes up to certain extent. In face of clutter, noise and variation in pose and illumination, object appearance can be disambiguated by the position of objects that real world scenes often exhibit. An example of this situation is presented in Figure 1. Information about typical con?gurations of objects in a scene has been stud- ied in psychology puter vision for years, in order to understand its e?ects in visual search, localization and recognition performance [2–4, 19, 23]. al.[4] proposed ?ve di?erent classes of relations between