Authors : V. Premchandran and P. Poongodi
Abstract: We propose a novel method for combined de-blurring and depth estimation in a variation framework. Our goal is to retrieve a de-blurred image and the corresponding depth map by only using information from a single blurred image. In many practical applications, for example for defocus or camera motion, the shape of the blurring remains constant but only its spatial extent is varying. A new blurring model based on this assumption in which the depth is included explicitly, is introduced. Depth is related to the spatial extent of the blur by assuming that the stronger an object is blurred, the closer it is to the camera. We also present a reduction of the model to one-dimensional blurring, as it is present for example in motion blur. The de-blurred image and the depth map is then estimated variation by means of gradient descent, where a novel partial equation occurs. Our method is evaluated experimentally and the results that the combination of de-blurring and depth estimation from just one image is possible.
V. Premchandran and P. Poongodi, 2016. Enhanced Image Processing for Blurred Images Using GA Based Image Extraction. Asian Journal of Information Technology, 15: 608-613.