Fast image reverse filters through fixed point and gradient descent acceleration

TL;DR

This paper enhances reverse image filtering techniques by applying acceleration methods to fixed point and gradient descent iterations, significantly improving convergence speed in estimating original images from black box filters.

Contribution

It introduces acceleration techniques for fixed point and gradient descent methods in reverse image filtering, improving their efficiency and convergence speed.

Findings

Acceleration methods significantly speed up convergence.

Fixed point and gradient descent approaches are effectively enhanced.

Experimental results demonstrate improved performance over existing methods.

Abstract

In this paper, we study the problem of reverse image filtering. An image filter denoted g(.), which is available as a black box, produces an observation b = g(x) when provided with an input x. The problem is to estimate the original input signal x from the black box filter g(.) and the observation b. We study and re-develop state-of-the-art methods from two points of view, fixed point iteration and gradient descent. We also explore the application of acceleration techniques for the two types of iterations. Through extensive experiments and comparison, we show that acceleration methods for both fixed point iteration and gradient descent help to speed up the convergence of state-of-the-art methods.