Restoration of multichannel images with limited a priori information using multichannel parallel-like adaptive filters

We introduce a heuristic method for adaptive multichannel restoration of degraded multispectral images when there is limited knowledge about the undegraded images and the noise. The proposed structure aims at reversing the action of the blur function on the image. The system uses number of channels equal to the number of multispectral images L. Each channel consists of L subchannel structures of two parallel adaptive filters with different cutoff frequencies and different DC (zero frequency) gain. The overall structure in each subchannel is equivalent to a filter, which has a combined LP-BP like characteristics. The resulting system has unity gain at the DC, which preserves the image local characteristics. The filter has high gain in the middle (band pass) frequency range, which produces an output image with restored edges. The filter structure gain at the high frequency range is small which reduces the high frequency noise. The adaptive filters are considered for the implementation of the proposed method. This is because they have many desirable properties and able to track the variations in the image characteristics. Multiple channel restoration results are presented and compared with single channel restoration. The results show that the proposed multiple channel restoration is very effective in restoring image details and reduces the noise amplification. Also the fine details are enhanced and preserved.