Improvement of interpolated color filter array image using modified mean-removed classified vector quantization

Authors:
Jim Z. C. Lai;Yi-Ching Liaw
Affiliations:
Department of Computer Science, National Taiwan Ocean University, Keelung, Taiwan 202, R.O.C.;Department of Optical Communications and Networking Technologies, Computer and Communications Research Labs, Room 722, Bldg. 51, No. 195, Chutung, ITRI Hsinchu, Taiwan 310, R.O.C.
Venue:
Pattern Recognition Letters
Year:
2005

Citing 2
Cited 2

New edge-directed interpolation

IEEE Transactions on Image Processing
Subpixel edge localization and the interpolation of still images

IEEE Transactions on Image Processing

Reduction of JPEG compression artifacts by kernel regression and probabilistic self-organizing maps

IWANN'11 Proceedings of the 11th international conference on Artificial neural networks conference on Advances in computational intelligence - Volume Part II
Adaptive kernel regression and probabilistic self-organizing maps for JPEG image deblocking

Neurocomputing

Quantified Score

Hi-index	0.10

Visualization

Abstract

To reduce the cost of digital cameras, the color filter array (CFA) is usually coated upon a single-chip image sensor. Each pixel of the image sensor can sense only one of the R, G, B colors under CFA. The two missing colors of a pixel are estimated using the CFA interpolation algorithm. The interpolation algorithm may generate color distortion and degrade the quality of interpolated images. To improve the quality of interpolated images, a modified mean-removed classified vector quantization algorithm is proposed to reduce the estimation error of interpolated colors. The algorithm extends and modifies vector quantization to discover the relationships between the interpolated images and their corresponding original versions using the information from CFA images. The discovered relationships are stored in codebooks and are used to improve the quality of images interpolated by the existing CFA interpolation algorithms. The experiments reveal that the proposed algorithm can improve the quality of images interpolated by the best CFA interpolation algorithm as far as we know. In terms of PSNR, the average PSNR improvements of R, G, and B channels are 0.89, 0.71, and 0.74dB, respectively.