DOI: 10.5176/978-981-08-8113-9_ITA2011-8
Authors: Kosuke Bannai, Kazuyuki Narisawa and Ayumi Shinohara
Abstract:
We investigate the usefulness of lossy compression-based distance for a similarity measure. If two objects are similar, the joint compressed size of them is much smaller than the sum of the two individual compressed objects. It is well known that the normalized compression distance (NCD) can be used for the similarity measure with lossless compressors [8]. It is based on the Kolmogorov complexity. In this paper, we propose a normalized lossy compression distance (NLCD) for lossy compressors. Moreover, we show the performance of NLCD for image data with the fractal image compression as lossy compressor.
Keywords: Kolmogorov Complexity; Lossy Compresssion; Normalized Compression Distance; Lossy; Image Retrieval; Normalized Lossy Compression Distance; Similarity
Updating...