A BENCHMARK FOR EDGE-PRESERVING IMAGE SMOOTHING

Abstract

Edge-preserving image smoothing is an important step for many low-level vision problems. Though many algorithms have been proposed, there are several difficulties hindering its further development. First, most existing algorithms cannot perform well on a wide range of image contents using a single parameter setting. Second, the performance evaluation of edge preserving image smoothing remains subjective, and their lacks a widely accepted datasets to objectively compare the different algorithms. To address these issues and further advance the state of the art, in this work we propose a benchmark for edge-preserving image smoothing. This benchmark includes an image dataset with ground truth image smoothing results as well as baseline algorithms that can generate competitive edge preserving smoothing results for a wide range of image contents. The established dataset contains 500 training and testing images with a number of representative visual object categories, while the baseline methods in our benchmark are built upon representative deep convolutional network architectures, on top of which we design novel loss functions well suited for edge-preserving image smoothing. The trained deep networks run faster than most state-of-the-art smoothing algorithms with leading smoothing results both qualitatively and quantitatively. The benchmark will be made publicly accessible.

Existing System

Edge-avoiding Wavelet (EAW), Rolling Guidance Filter (RGF), Weighted Median Filter.

Proposed System

In this paper a benchmark for edge-preserving image smoothing is proposed. This benchmark includes an image dataset with “groundtruth” image smoothing results as well as baseline models that are capable of generating reasonable edge-preserving smoothing results for a wide range of image contents. The image dataset contains 500 training and testing images with a number of visual object categories, including humans, animals, plants, indoor scenes, landscapes and vehicles. The groundtruth smoothing results in our dataset are not directly generated by handcraft approaches, but manually chosen from results generated by existing state-of-the-art edge-preserving smoothing algorithms. This is justified by two reasons. Specifically, very deep convolutional networks (VDCNN) and deep residual networks (ResNet) is used. On top of these network architectures, we design novel loss functions well suited for edge-preserving image smoothing. The deep networks trained over our dataset run faster than most state-of-the-art edge-preserving smoothing algorithms, while the smoothing performance of our ResNet based model outperforms these algorithms both qualitatively and quantitatively.

Architecture

BLOCK DIAGRAM