Image Bit-Depth Manipulation using Distance Fields

Bügel, Niels (2023) Image Bit-Depth Manipulation using Distance Fields. Master's Thesis / Essay, Computing Science.

Preview	Text mCS_2023_BugelNA.pdf Download (59MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (130kB)

Abstract

With the increase in popularity of high dynamic range (HDR) images, there is also an increasing demand for methods that can increase the bit-depth of existing low dynamic range images. Additionally, HDR images come with larger storage requirements, which is often undesirable. We present a method that can be utilised for both bit-depth expansion and bit-depth compression. To enhance the bit depth, we interpret the input image as a 3D surface. Our method makes use of signed distance field blending to smooth this surface, allowing us to remove visible contours while maintaining edge data. We make use of this technique to construct the decompression step in a new lossy bit-depth compression scheme. During the compression step, we remove the grey levels that are not expected to contribute significantly to the reconstruction process. The evaluation of the scheme is done using NVIDIA's FLIP algorithm; a state-of-the-art metric for assessing perceived differences between images. For the implementation and evaluation of our method, we developed a new tool called NITRO. This is a powerful node editor that allows for the construction of complex image-processing routines via a visual graph representation. Using this tool, we showcase and evaluate the performance of the compression scheme on various images.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Kosinka, J. and Da Silva, D.J.
Degree programme:	Computing Science
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	19 Jul 2023 07:12
Last Modified:	19 Jul 2023 07:12
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/30742

Actions (login required)

View Item