Lens Distortion Calibration Using Level Sets

Can lens distortion be calibrated from an image of a smooth, textureless Lambertian surface?

“…The image of Fig. 3(b) is also used to estimate the radial dis- tortion coefficient by two well-known methods: namely the well-known line straightness method [2] and a variational approach for lens distortion calibration based on level sets [5].…”

“…Such distributions were derived in [11] to solve for the illuminant direction. Here we assume the same distributions in order to estimate the expected values of the ratios in (5). The assumed-independent distributions f α and f β of α and β, respectively, are taken [11] …”

“…The second category of non-metric methods does not rely on calibration objects of known structure. The majority of such methods rely on the presence of straight lines in the scene [2,3,4,5,6]. These methods are easier to use, but have the possible expense of lower accuracy or stability of results.…”

A PDE Method to Segment Image Linear Objects with Application to Lens Distortion Removal

“…One may notice some artifacts (left intentionally) with the undistorted images due to the inverse mapping of the distortion model in (28), which can be fairly fixed, if desired, by doing some post-processing. Further experiments on synthetic data, [29], have shown that the accuracy of our proposed method remains within 0.1 pixels up to a high noise level of 35 σ ≅ . …”

What Can an Image of a Smooth Lambertian Surface Tell About Camera Nonlinearity?

“…There is also another category of non-metric methods that do not need calibration objects of known structure. These methods depend largely on extracting specific image features, e.g., lines [18], [19], [20], [21], or spheres' occluding contours [22], [17], [23]. As such, their success depends on the reliability of localizing these features, which may require some user guidance or supervision.…”