inlier

An inlier is an observation that lies within a data set's expected structure under a chosen model or distribution. In most contexts, an inlier agrees with the model and is treated as part of the underlying pattern; an outlier is an observation that deviates significantly from the model.

In statistics and data analysis, inliers are points with small residuals when data are fitted by a model; for example, in linear regression, an inlier is a point whose actual value is close to the predicted value. In robust estimation and model fitting, algorithms such as RANSAC identify inliers as data points whose distance to the hypothesized model is below a threshold; the inlier set is then used to refine the model.

In computer vision and photogrammetry, the term is used when fitting geometric relations such as a homography or a fundamental matrix to matched features. Points whose coordinates satisfy the geometric constraint within a tolerance are labeled inliers; points that fail the constraint are considered outliers.

In practice, identifying inliers helps separate true data structure from noise or erroneous measurements. However, the choice of threshold can affect results, and datasets may contain multiple underlying structures that require multiple inlier groups or models. Overall, the concept of inliers supports robust data analysis by focusing on measurements that conform to the assumed model.