I have accomplished my BSc and MSc in Mathematics and Physics at the University of Warwick in the United Kingdom and I’m a researcher at the world leading company in lightfield camera technology Raytrix.
My current and future research interests include respectively analysis and modelling of point spread functions for lightfield cameras and will developing an elaborate mathematical distortion model for enhanced metric calibration.
The point spread function (PSF) of an optical system gives fundamental insight into the imaging performance of a camera. So modelling and understanding this optical property of the imaging system is significantly advantageous in a number of ways.
One major complication however is that in a lightfield camera, there is not a single PSF, but rather an array of them formed behind the microlenses imaging the object point relayed through the main lens.
Having an accurate PSF simulation model at hand allows for faster computational imaging and scene rendering for synthetic data, (as intensity images form from a convolution of the PSF with the object profile). From this, one can then test the performance of depth estimation algorithms of pixel matching and how these change with the optical setup. Thus we can ultimately optimise the system’s configuration to achieve the most accurate depth maps.
Furthermore, by knowing the depth dependent PSF, we can construct our lightfield cameras better. This is because with knowledge of the PSF, we can attain the relationship between the intensity PSF spot size at the sensor with respect to the microlens array (MLA) to sensor distance.
As lightfield camera already provides us with a depth map of our scene, then for a corresponding object depth, we can determine an initial estimate on the defocus parameter. As a result, one can then apply an iterative algorithm which alternates between estimating depth from disparity and deconvolution for image restoration of lightfield images.
As a result we obtain enhanced lateral resolution in our final microimages and therefore improved depth/3D scene reconstruction.
Mehdi Daniel Ardebili