The spatial and chromatic dimensions of digital color image information exhibit unique interrelationships that invite new, color-image-specific, processing strategies. A quantitative exploration of these interrelationships was performed. The resulting data revealed key traits that led to two new methods of color image enhancement.

The first is a method of color image contrast enhancement that exploits the existence of high-pass spatial energy in certain chromatic color components. We present a new, spatially adaptive approach that acknowledges the spatially varying nature of cross-component correspondences. This new approach is suitable for use in any color space.

The second is a method of color image denoising that exploits the unique correspondences of polychromatic multiscale edges at fine scales of analysis. The multiscale edges are derived using wavelet methods. This approach preserves image details and noticeably outperforms wavelet thresholding methods of denoising in images containing natural foliage.

Images below show the performance of the proposed methods:

   
a) Noisy Image                                          b) De-Noised Image

Publications:

1. Bruce A. Thomas and Jeffrey J. Rodriguez, "Wavelet-Based Color Image Denoising," in Proc. 2000 IEEE Intl. Conf. on Image Processing, Vancouver, B.C., Canada, Sept. 10-13, 2000, vol. 2, pp. 804-807. [ PDF ]
   
   
2. Bruce A. Thomas and Jeffrey J. Rodriguez, "Practical Implementation of Multirate Convolution for Multiresolution Image Processing," in Proc. 1998 IEEE Southwest Symp. on Image Analysis and Interpretation, Tucson, AZ, April 6-7, 1998, pp. 217-222. [ PDF ]
   
   
3. Bruce A. Thomas, Robin N. Strickland, and Jeffrey J. Rodriguez, "Color Image Enhancement Using Spatially Adaptive Saturation Feedback," in Proc. IEEE Intl. Conf. on Image Processing, Santa Barbara, CA, Oct. 26-29, 1997, vol. 3, pp. [ PDF ] 30-33.