Abstract
Chain codes compactly represent raster curves, but there is still a lot of room for improvement by means of data compression. Several statistics-based chain code compression techniques assign shorter extra codes to frequent pairs of consecutive symbols. Here we systematically extend this concept to patterns of up to six symbols. A curve may be represented by any of the exponentially many overlapped chains of codes, and the dynamic programming approach is proposed to determine the optimal chain. We also propose utilization of multiple averaged hard coded pseudo-statistical models, since the exact statistical models of individual curves are often huge, and they can also significantly differ from each other. A competitive compression efficiency is assured in this manner and, as a pleasant side effect, this efficiency is less affected by the curve’s shape, rasterization algorithm, noise, and image resolution, than in other contemporary methods, which surprisingly do not pay any attention to this problem at all.
Highlights
206 Informatica 45 (2021) 205–212 based approaches were achieved by introducing extra codes for frequent pairs of primitives [7, 8], or by utilization of multiple statistical models in so-called context-based approaches [1], where the statistical model for coding a considered symbol is conditioned on the context of M previous symbols
We introduce a new statistics-based approach where extra codes may represent patterns of up to k = O(1) symbols
We compare some typical results of the proposed method and some state-of-the-art (SOTA) chain code compression methods. 3OT, VCC, compressed DDCC (C_DDCC), and three variants of MTFT+ARLE (Move-To-Front Transform + Adaptive Run-Length Encoding) [11], i.e., MTFT+ARLE VCC, MTFT+ARLE 3OT, and MTFT+ARLE NAD [11] were used in the tests
Summary
206 Informatica 45 (2021) 205–212 based approaches were achieved by introducing extra codes for frequent pairs of primitives [7, 8], or by utilization of multiple statistical models in so-called context-based approaches [1], where the statistical model for coding a considered symbol is conditioned on the context of M (typically 1 or 2) previous symbols. Non-statistical approaches perform various string transformations [11, 12], e.g. BurrowsWheeler Transform (BWT) and/or Move-To-Front Transform (MTFT) to increase the number of 0’s and prepare the data for efficient run-length encoding (RLE) and/or binary arithmetic coding (BAC). Impacts of the curve shape, image resolution, rasterization algorithm, noise, and geometric transformations on the compression ratio are significantly reduced in comparison to other contemporary methods. This topic has been so far addressed indirectly within the context-based approaches and, partially, in the non-statistical approaches, while it was completely neglected in other related works.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
