Available Technology
A new framework for wavelet-based analysis and processing of color filter array images
Technology:
Method of de-mosaicing spatially sampled image data
Markets Addressed
The invention describes a new approach to de-mosaicing of spatially sampled image data observed through a color filter array (CFA). The invention employs properties of Smith-Barnwell filterbanks to exploit the correlation of color components in order to reconstruct a sub-sampled image. In particular, the invention admits a new framework for wavelet-based CFA image de-noising and de-mosaicing methods, which in turn enables the application of existing wavelet-based image de-noising techniques directly to sparsely sampled data. This capability is important because various noise sources inherent to the charge-coupled device or other imaging technique employed must be taken into account in practice; any noise reduction procedure should ideally take place prior to de-mosaicing (both to improve interpolation results and to avoid introducing additional correlation structure into the noise). While earlier work has been focused primarily on de-mosaicing prior to de-noising, this new method suggests a natural way to perform wavelet-based de-noising and de-mosaicing together in tandem. Results indicate that this new method performs on par with the state-of-the-art method for far lower computational cost, and provides a versatile, effective, and low-complexity solution to the problem of interpolating color filter array data observed in noise.
Applications include:
-De-mosaicing of color image data for digital still cameras and digital video cameras
-Color image processing
Innovations and Advantages
In digital imaging applications, data are typically obtained via a spatial sub-sampling procedure implemented as a color filter array (CFA). CFA assigns a separate primary color to each pixel by placing a filter of that color over the pixel. The most well known CFA is the Bayer pattern which uses a checkerboard pattern with alternating rows of filters. The Bayer filter has twice as many green pixels as red or blue and takes advantage of the human eye's tendency to see green luminance as the strongest influence in defining image quality.
The term de-mosaicing refers to the inverse problem of reconstructing an image from the partial raw data received from the color-filtered image sensors internal to many digital cameras in the form of a matrix of colored pixels. It is well known that the most efficient solution to this inverse problem produces unacceptable visual distortions and artifacts.
Aside from the spatial under-sampling inherent in the Bayer pattern, this phenomenon can be attributed to the observation that values of the color triple exhibit significant correlation, particularly at high spatial frequencies – such content often signifies the presence of edges, whereas low-frequency information contributes to distinctly perceived color content. As such, most de-mosaicing algorithms attempt to make use of this correlation structure in the spatial frequency domain.
The advantage of this invention is that it admits a new framework for wavelet-based CFA image de-noising and de-mosaicing methods, which in turn enables the application of existing wavelet-based image de-noising techniques directly to sparsely sampled data.
Additional Information
Tweet
Inventor(s):
Hirakawa, Keigo
Meng, Xiao-Li
Wolfe, Patrick J.
Categories:
For further information, please contact:
Sam Liss, Director of Business Development
(617) 495-4371
Reference Harvard Case #2838