Computer vision methods for mobile imaging and 3D reconstruction

Thesis event information

Date and time of the thesis defence

Place of the thesis defence

Linnanmaa, L10, remote link

Topic of the dissertation

Computer vision methods for mobile imaging and 3D reconstruction

Doctoral candidate

Master of Science (Technology) Janne Mustaniemi

Faculty and unit

University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, Center for Machine Vision and Signal Analysis (CMVS)

Subject of study

Computer Science and Engineering


Professor Joni-Kristian Kämäräinen, Faculty of Information Technology and Communication Sciences, Tampere University


Professor Janne Heikkilä, Faculty of Information Technology and Electrical Engineering, University of Oulu

Add event to calendar

Computer vision methods for mobile imaging and 3D reconstruction

The imaging capabilities of smartphones have advanced considerably over the last years. In good imaging conditions, smartphone cameras can already compete with bulky and expensive DSLR cameras. Problems commonly occur when shooting in low light conditions at handheld. A small camera is able to collect a very limited amount of light which leads to noisy images. Moreover, the image may appear blurry if the camera or scene objects move during the image exposure. Motion blur not only degrades the visual quality but damages various computer vision applications, including image-based 3D reconstruction.

This thesis presents two image deblurring methods that utilize an inertial measurement unit (IMU) commonly found in smartphones. The IMU provides information about the motion of the device, which is valuable when removing motion blur. This thesis also investigates the problem of joint denoising and deblurring. Many devices can be programmed to capture rapid bursts of images with different exposure times. This work introduces a novel learning-based approach to recovering sharp and noise-free photographs from a pair of short and long exposure images.

Image-based 3D reconstruction is an essential problem in computer vision. The goal is to create a 3D model of the environment from a collection of images or a video. This process has a well-known limitation that the absolute scale of the reconstruction cannot be recovered using a single camera. This thesis presents an inertial-based scale estimation method that recovers the unknown scale factor. The method achieves state-of-the-art performance and can easily be integrated with existing 3D reconstruction software.

Multi-aperture cameras have become common in smartphones. The use of multiple camera units provides another way to improve image quality and camera features. This thesis explores the problem of parallax correction caused by each camera unit having a slightly different viewpoint. This work presents an image fusion algorithm for a particular multi-aperture camera where camera units have different color filters. The images are fused using a disparity map that is estimated while considering all images simultaneously. The approach is a feasible alternative to traditional cameras equipped with a Bayer filter.
Last updated: 1.3.2023