Real-Time Restoration for Motion-Blurred Color Images by Multichannel Combination Processing

A fast algorithm is proposed to restore the color images, degraded in a deterministic way by motion- blurred,by multichannel combination processing. Based on the color space conversion model, the point spread functions of the image motion on each plane of YUV422 image can be achieved.The U, V plane of YUV422 images are combined to form a newel plane which has the same size as Y plane. Then, the Y, UV plane are packed into a two dimensional complex array which can be transformed directly by FFT and result a composite frequency spectrum. Wiener filter is applied to filter Y, UV spectrums separated from the composite spectra. On Graphics Processing Unit (GPU) platform, the proposed algorithm can restore a 1024times1024 24-bit motion-blurred color image in 7.95 ms, and the PSNR of the restored image is 27.12 which reveals a good restoration performance.