期刊
IEEE TRANSACTIONS ON MULTIMEDIA
卷 23, 期 -, 页码 2902-2916出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TMM.2020.3019688
关键词
Feature extraction; Saliency detection; Object detection; Image reconstruction; Image segmentation; Image color analysis; Neural networks; Attention map; background subtraction; salient region; affinity matrix; convolution neural network
资金
- Hong Kong Research Grants Council [C1007-15G]
- Hong Kong Institute for Data Science [8730039]
By utilizing high-level features and affinity-based techniques, a precise salient region can be extracted effectively from noisy and cluttered backgrounds, controlling foreground and background information to enhance detection quality.
Most existing saliency methods measure fore- ground saliency by using the contrast of a foreground region to its local context, or boundary priors and spatial compactness. These methods are not powerful enough to extract a precise salient region from noisy and cluttered backgrounds. To evaluate the contrast of salient and background regions effectively, we consider high-level features from both supervised and unsupervised methods. We propose an affinity-based robust background subtraction technique and maximum attention map using a pre-trained convolution neural network. This affinity-based technique uses pixel similarities to propagate the values of salient pixels among foreground and background regions and their union. The salient pixel value controls the foreground and background information by using multiple pixel affinities. The maximum attention map is derived from the convolution neural network using features of the Pooling and Relu layers. This method can detect salient regions from images that have noisy and cluttered backgrounds. Our experimental results demonstrate the effectiveness of the proposed approach on six different saliency data sets and benchmarks and show that it improves the quality of detection beyond current saliency detection methods.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据