Fast approximate inverse filtering for reducing motion blur in liquid crystal display

This paper presents an approximate deblur filter for reducing motion blur that appears on liquid crystal displays due to their slow response time and hold-type property. The filter coefficients are derived from the regularized inverse of blur model, which are in the form of concatenation of two or more window functions (low pass filters) separated by the amount of motion. Hence, it is shown that the motion deblur filter for the current frame can be approximated as a weighted sum of two or more low pass filtered consecutive frames, which needs very simple computations. More specifically, each frame is filtered with a 2D lowpass filter, where its size and shape are controlled according to the magnitude and direction of motion vector. Then the output is simply obtained by subtracting previous frame from the current frame with some control weights. Experimental results show that the proposed method provides comparable performance to the existing frequency domain methods, while requiring much lower computational complexity.