Abstract
We would like to simulate and explore the deblurring of images with linear, one-dimensional blur. This blur is the result of the objectbeing in 1-D motion while the shutter is open. The effect is similar to convolving a boxcar filter and smearing the image in it's direction ofmoving. The boxcar filter has a “sinc” spectrum, which causes the high-frequency details to be lost. In addition, deconvolving a boxcaralso magnifies the high frequency noise. These together render the result of recovery undesirable.
By fluttering the shutter during the exposure with a random binary sequence, we broaden the smearing filter spectrum, which preserves morehigh frequency details in the blurred image. This produces a decent deconvolution result with less noise. We also exploited and compared theresult of deblurring in both time and frequency domain using both coded and uncoded shutter.