The invention relates to a method for determining a localization error (ε) of a point (P0) of a raw image, comprising the following steps: estimating the value of a statistical magnitude (G) characteristic of a probability law (D(X, Y, Z)) of at least one terrain coordinate (X, Y, Z) associated with the point (P0) of the raw image, using a probability law (D(θ1, . . . , θn)) with magnitudes (θ1, . . . , θn) depending on the exposure conditions of the raw image and a localization function, deduced from an exposure function and a terrain model and applied for the image coordinate point (P0) of the raw image; and deducing the localization error (ε) of the point (P0) of the raw image from the statistical magnitude (G).