Convolutional Neural Pyramid for Image Processing

TL;DR

This paper introduces a convolutional neural pyramid framework that efficiently enlarges receptive fields for various low-level vision tasks, enabling high-quality image processing with reduced computational cost.

Contribution

The proposed neural pyramid structure achieves large receptive fields without high computational costs, supporting diverse image processing applications with adaptive depth and progressive upsampling.

Findings

Enables quasi-realtime processing of VGA images

Improves performance in image restoration and enhancement tasks

Reduces computational complexity compared to traditional methods

Abstract

We propose a principled convolutional neural pyramid (CNP) framework for general low-level vision and image processing tasks. It is based on the essential finding that many applications require large receptive fields for structure understanding. But corresponding neural networks for regression either stack many layers or apply large kernels to achieve it, which is computationally very costly. Our pyramid structure can greatly enlarge the field while not sacrificing computation efficiency. Extra benefit includes adaptive network depth and progressive upsampling for quasi-realtime testing on VGA-size input. Our method profits a broad set of applications, such as depth/RGB image restoration, completion, noise/artifact removal, edge refinement, image filtering, image enhancement and colorization.