Defocus Map Detection Using a Single Image

Abstract:

The estimation of blurred regions is an important stage in several computer vision applications. In this paper an efficient training-free detector of local blurriness based on edge features is presented. Due to the intrinsic sparsity of edges in natural images a blur map is creating by using an approach based on the heat diffusion principle. A 2D point discrete Poisson solver is concatenated with a guided filter stage in order to create the blurring map. Experiments with images from two publicly available datasets validate the proposed method.

Año de publicación:

2016

Keywords:

• blur map
• out of focus
• Poisson solver

Fuente:

scopus

google