Homomorphic Filtering Approach for Narrow Band Images Enhancement

The application of narrow band imaging (NBI) in endoscopy improves the accuracy in diagnosis, including reduction in cost and time. It allows better definition between the epithelial surface and the adjacent vascular net via the application of narrow-bandwidth filters to standard white-light endoscopy. The fact, that only selected bandwidths are emitted results in a lower brightness, which requires special high –sensitive dual mode CCDs. In the paper is presented an approach for NB images enhancement using improved homomorphic filtering. Traditional homomorphic filtering is based on Fourier transform, which has higher resolution in frequency domain and lower resolution in spatial domain. The proposed algorithm includes homomorphic filtering based on wavelet packet transform as first step of improving, and contrast limited adaptive histogram equalization (CLAHE) as the second improving step to adjust its whole luminousness. An adaptive threshold estimation method for image denoising in the wavelet domain is used, for modeling of detail sub-band coefficients. An effective algorithm was treatment by optimization of some parameters of the CLAHE and wavelet transformation on the base of objective quantitative estimation parameters. The main goal is noise reduction and improving the quality of the diagnostic images. Some results of the experiments with colonoscopy images are presented, which were made by computer simulation in MATLAB environment.