In this talk, I will present my work on image deblurring and denoising with the
help of auxiliary information. Modern mobile devices are commonly equipped with
an inertial measurement unit (IMU) providing an initial estimate of the motion
blur. We show that by utilizing IMU readings, the deblurring can be performed
realtime. Furthermore, the state-of-the-art deblurring results can be achieved
with a deep neural network that utilizes both the IMU readings and image data.
Finally, we capture pairs of short and long exposure images in rapid
By jointly denoising and deblurring, we can recover the sharp and noise-free
image even if the scene contains moving objects. This approach also allows the
capture of HDR images with a moving camera.