Estimation of Blur and Depth_Map of a De-focused Image by Sparsity using Gauss Markov Random Field Convex-Prior

Latha H N, H N Poornima

Abstract: In this work, we propose a new method for blur-map and depth estimation from de-focused observations using just noticeable blur (JNB) [1] method. Using JNB, we find the blur-map and then estimate the depth of the image in the depth from de-focus setting. We use a novel regularization based optimization framework, wherein we assume the blur-map as Gauss Markov random field. We initially obtain robust estimates of the blur-map then depth of the scene using a convex prior [2]. We show that JNB and clear dictionaries are not replaceable when conducting sparse patch reconstruction. We also show that the estimated blur-map which is utilized for efficient restoration of latent image by de-blurring.

Keywords: Space-variant Blur-map, Just Noticeable Blur, Gradient Descent, GMRF, Convexity

How to Cite?: Latha H N, H N Poornima, "Estimation of Blur and Depth_Map of a De-focused Image by Sparsity using Gauss Markov Random Field Convex-Prior", Volume 8 Issue 11, November 2019, International Journal of Science and Research (IJSR), Pages: 728-733, https://www.ijsr.net/getabstract.php?paperid=ART20202687, DOI: https://dx.doi.org/10.21275/ART20202687

Download Citation: APA | MLA | BibTeX | EndNote | RefMan