Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Blind Image Deblurring Using Spatial Stimuli Gradient Based Regularization (BDSSGR)
1Software Engineering Department, University of Engineering and Technology, Taxila, Pakistan.
2Department of Information Engineering Technology, National Skills University Islamabad, Pakistan.
ABSTRACT
Blind image deblurring is considered as an ill pose problem. Several techniques have been proposed to cater this ill posed problem. However, the recovered image is still found to be blurred and have low contrast with sparse edges. In this paper, an efficient technique based on combination of spatial stimuli gradient model and normalized regularization is proposed named as BDSSGR. The BDSSGR is mainly divided into three steps. Firstly, the spatial stimuli gradient Model (SSGM) is used to restore the salient edges. Which is obtained by using the amount of brightness perceived and similarity among neighboring pixels. In the second step, the scale invariant regularization is applied to estimate the kernel via alternate steps between updating the unknown image and kernel. Lastly, a non-blind deconvolution algorithm is employed to restore the sharp image. The evaluation validates that BDSSGR achieves better results on synthetic and real word blur image datasets as compared to other state-of-the-art techniques.
Keywords: Image deblurring, Edge detection, Spatial stimuli gradient model, Kernel estimation, Regularization.
Download PDF
1Software Engineering Department, University of Engineering and Technology, Taxila, Pakistan.
2Department of Information Engineering Technology, National Skills University Islamabad, Pakistan.
ABSTRACT
Blind image deblurring is considered as an ill pose problem. Several techniques have been proposed to cater this ill posed problem. However, the recovered image is still found to be blurred and have low contrast with sparse edges. In this paper, an efficient technique based on combination of spatial stimuli gradient model and normalized regularization is proposed named as BDSSGR. The BDSSGR is mainly divided into three steps. Firstly, the spatial stimuli gradient Model (SSGM) is used to restore the salient edges. Which is obtained by using the amount of brightness perceived and similarity among neighboring pixels. In the second step, the scale invariant regularization is applied to estimate the kernel via alternate steps between updating the unknown image and kernel. Lastly, a non-blind deconvolution algorithm is employed to restore the sharp image. The evaluation validates that BDSSGR achieves better results on synthetic and real word blur image datasets as compared to other state-of-the-art techniques.
Keywords: Image deblurring, Edge detection, Spatial stimuli gradient model, Kernel estimation, Regularization.
Download PDF