Monocular 3D metric scale reconstruction using depth from defocus and image velocity

Eliminating Scale Drift in Monocular SLAM Using Depth From Defocus

“…However, measuring σ i does not work well as expected on features other than those with sharp edges due to the blur texture ambiguity [8]. In our previous work [9], we found that one of the main causes was the difference of the contrast between the ROIs and demonstrated that this error could be expressed by the following equation:…”

“…Third, given a single image, it cannot be differentiated whether a blur is caused by defocus or texture [22]. In our previous work [9], we demonstrated that the blur due to texture could be represented using a constant correction factor and an EKF framework was able to produce accurate metric reconstruction.…”

“…In our previous work [9], we demonstrated that the amount of defocus blur cannot be estimated correctly on complex texture such as letters, and therefore constraints of Eq. (5) no longer hold in these situations.…”

“…It relies on the amount of defocus blur which is related to the distance to the scene [8]. In our previous work [9], we demonstrated that combination of DfD and image velocity in an extended Kalman filter (EKF) framework was able to estimate the metric scale of an environment accurately. In this paper, we leverage this work to integrate DfD with monocular SLAM systems.…”