Aliasing artifacts reduction with subband signal analysis for demosaicked images

Ji Yong Kwon, Sang Wook Park, Min Kyu Park, Moon Gi Kang

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Demosaicking is a process used to estimate missing color values from the subsampled color filter array (CFA) image to reduce the cost and volume of a digital still camera. However, by sampling theory, it is known that subsampling a signal causes overlaps of signals in the frequency domain, which is known as aliasing. Most current demosaicking processes cannot completely solve aliasing problem resulting in aliasing artifacts such as false colors and zipper effects. In this paper, we propose an algorithm to remove these aliasing artifacts in demosaicked color images. A luminance image with minimum aliasing is obtained from the CFA image by using a low-pass kernel with cutoff frequencies determined by an approximate model for the Fourier spectrum. An aliasing map is computed by analyzing subband signals of the CFA image based on the high correlation of the high-frequencies of the luminance and color channels. Then, a least squares of the luminance acquisition processes is used to design a cost function with the aliasing map to remove the aliasing artifacts. The experiments demonstrate that the proposed algorithm sufficiently removes aliasing artifacts and improves the quality of the color images.

Original languageEnglish
Pages (from-to)115-128
Number of pages14
JournalDigital Signal Processing: A Review Journal
Volume59
DOIs
Publication statusPublished - 2016 Dec 1

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Statistics, Probability and Uncertainty
  • Computational Theory and Mathematics
  • Electrical and Electronic Engineering
  • Artificial Intelligence
  • Applied Mathematics

Fingerprint Dive into the research topics of 'Aliasing artifacts reduction with subband signal analysis for demosaicked images'. Together they form a unique fingerprint.

  • Cite this