POCS-DSR

POCS based Depth Super Resolution

 Partecipanti

Mappa

 Word cloud

Esplora la "nuvola delle parole (Word Cloud) per avere un'idea di massima del progetto.

 Obiettivo del progetto (Objective)

'The objective of this proposal is to increase the range resolution of time-of-flight (ToF) cameras by utilizing correlation among multi-frame data through projection onto convex sets (POCS) technique. Our work aims to produce a super-resolution depth map and reduce the depth error. A calibration based technique is proposed to form constraint sets that will compensate for spatially varying depth errors within the work volume. Our proposed framework incorporates adaptive kernels along with novel constraint sets to form high resolution depth maps. ToF cameras have the capability of producing real-time depth maps, reaching nearly up to 100fps. They are also compact in the sense that no moving gimbals or extra parts such as mirros and scanners are required for data acquisition. One of the application areas in which ToF cameras are used is 3D scene modeling which has matured to great extent during last decade. In addition to that, high-end automobile industry, production lines, and biomedical inspection are some of the areas in which range sensing is applied. The purpose of our proposal is to provide a way for ToF cameras to realize their potential for range sensing by increasing their depth resolution.'

Introduzione (Teaser)

EU-funded scientists are seeking to further develop time-of-flight (TOF) camera potential for range sensing by increasing their depth resolution.

Descrizione progetto (Article)

Use of TOF cameras for three-dimensional scene modelling has matured to a great extent over the last decade. Such range sensing can also be applied to high-end automobile industry, production lines and biomedical imaging. In the EU-funded project 'POCS based depth super resolution' (POCS-DSR), researchers are developing innovative algorithms in a bid to increase depth resolution while reducing depth error in a range-imaging camera.

Through a projection-onto-convex-sets (POCS) technique, the project is using multi-frame data correlation to produce a super-resolution depth map. A calibration-based technique is proposed to form constraint sets that should compensate for spatially varying depth errors within the work volume.

Researchers used an alignment technique to register real-time images on a digital signal processor platform. They devised a multi-exposure depth enhancement algorithm to capture and fuse valuable depth information in the scene using alternating exposure durations. Furthermore, parallel image reconstruction algorithms were developed for POCS.

Through innovative and efficient algorithms, researchers successfully met the real-time image processing requirements for several applications in the host organisation. Project activities were published in several conference papers at local and international level.