Underwater image enhancement using adaptive filtering for enhanced sift-based image matching

Pulung Nurtantio Andono, I. Ketut Eddy Purnama, Mochamad Hariadi

Research output: Contribution to journalArticlepeer-review

13 Citations (Scopus)


Success of scale-invariant feature transform (SIFT) image registration is limited when attempted on camera footage taken under water. This is, largely due to the poor image quality inherent to imaging in aquatic environments. In this research we aim to overcome this shortcoming using a new method of pre-processing of true-color imagery taken under water based on the Contrast Limited Adaptive Histogram image Equalization (CLAHE) algorithm. CLAHE assumes that the distribution function of the pixel intensity values of an underwater-recorded image is dominated by Rayleigh scattering, and that the noise can be removed as a function hereof. Results showed that after applying the CLAHE image enhancement method registration success of SIFT increased by 41% compared to reference method (a straightforward contrast stretching enhancement). From the ANOVA results follows that the null hypothesis can be rejected and concluded that there is a significant difference among SIFT without enhancement, SIFT with contrast stretching and SIFT with CLAHE-Rayleigh at 5% level of significance (ANOVA, F=23.41, df=2, p-value=7.34e-09). The finding concludes that CLAHE-Rayleigh is better compared to contrast stretching. As a follow-up study, CLAHE-Rayleigh should be compared with other image enhancement techniques (or a combination of techniques 'hybrids') to assess their relative impact on the success of SIFT-based image matching of underwater photography.

Original languageEnglish
Pages (from-to)273-280
Number of pages8
JournalJournal of Theoretical and Applied Information Technology
Issue number3
Publication statusPublished - Jun 2013
Externally publishedYes


  • Image enhancement
  • Image registration
  • SIFT
  • Underwater image processing


Dive into the research topics of 'Underwater image enhancement using adaptive filtering for enhanced sift-based image matching'. Together they form a unique fingerprint.

Cite this