☆ 4.5 Article

Conditional Random Field based salient proposal set generation and its application in content aware seam carving

SIGNAL PROCESSING-IMAGE COMMUNICATION (2020)

Journal

SIGNAL PROCESSING-IMAGE COMMUNICATION

Volume 87, Issue -, Pages -

Publisher

ELSEVIER

DOI: 10.1016/j.image.2020.115890

Keywords

Structured prediction; Edge classification; Salient edge density based scoring; Bayesian probabilistic edge map; Salient region preserving seam carving

Categories

Engineering, Electrical & Electronic

Ask authors/readers for more resources

Protocol

Community support

Reagent

Community support

Abstract

Most of the existing generic object localization algorithms usually give the plausible object locations without taking into consideration the saliency ordering of the proposal set. This paper presents a novel object proposal generation which ranks the key objects according to their saliency score in the proposal pool. First, we formulate a Bayesian framework for generating a probabilistic edgemap which is used to assign a saliency value to the edgelets. A conditional random field is then learnt for edge-labeling by effectively combining the edge features with the relative spatial layout of the edge segments. Lastly, we propose an objectness score for the generated proposal set by analyzing the salient object edge density completely lying within the candidate boxes. Extensive experiments on the benchmark PASCAL VOC 2007 and 2012 datasets demonstrate that the proposed method provides competitive performance against popular generic object detection techniques while using fewer number of proposals. Additionally, we demonstrate the applicability of the generated proposal set for content aware image retargeting.

Authors

I am an author on this paper

Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.5

Not enough ratings

Secondary Ratings

Novelty

-

Significance

-

Scientific rigor

-

Rate this paper

Recommended

Article Computer Science, Software Engineering

Analysis of seam carving technique: limitations, improvements and possible solutions

Ankit Garg, Anuj Kumar Singh

Summary: This paper presents an improved technique for image retargeting, which uses the gradient energy map generation method to safeguard the prominent regions of the image. The efficiency of the technique is validated through objective and subjective image quality assessment.

VISUAL COMPUTER (2023)

Add to Collection

Article Computer Science, Information Systems

Improved seam carving for structure preservation using efficient energy function

Ankit Garg, Anand Nayyar, Anuj Kumar Singh

Summary: This paper proposes an improved seam carving technique to overcome the limitations of traditional techniques by restricting the intersection of optimal seam points on line structures. Through subjective and objective evaluations, the effectiveness and high similarity of the proposed technique have been demonstrated.

MULTIMEDIA TOOLS AND APPLICATIONS (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

Motion Context guided Edge-preserving network for video salient object detection

Kan Huang, Chunwei Tian, Zhijing Xu, Nannan Li, Jerry Chun-Wei Lin

Summary: This paper proposes a Motion Context guided Edge-preserving network (MCE-Net) model for video salient object detection. MCE-Net can generate temporally consistent salient edges and refine the salient object regions completely and uniformly. Experimental results demonstrate the superior performance of the proposed method on five widely-used benchmarks.

EXPERT SYSTEMS WITH APPLICATIONS (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Hierarchical and Interactive Refinement Network for Edge-Preserving Salient Object Detection

Sanping Zhou, Jinjun Wang, Le Wang, Jimuyang Zhang, Fei Wang, Dong Huang, Nanning Zheng

Summary: Salient object detection has seen rapid development with the rise of Deep Neural Networks, with HIRN proposed in this paper as a novel and effective approach to preserve edge structures and improve accuracy in detection.

IEEE TRANSACTIONS ON IMAGE PROCESSING (2021)

Add to Collection

Article Computer Science, Hardware & Architecture

Privacy-preserving edge caching: A probabilistic approach

Seyedeh Bahereh Hassanpour, Ahmad Khonsari, Masoumeh Moradian, Seyed Pooya Shariatpanahi

Summary: Edge caching reduces access delay but increases leakage probability. The study proposes a chunk-based joint probabilistic caching approach to mislead eavesdropping adversaries and optimize cache placements. The proposed scalable approach outperforms existing methods in terms of privacy and communication cost.

COMPUTER NETWORKS (2023)

Add to Collection

Article Computer Science, Information Systems

Performance analysis of seam diversion based image retargeting technique based on edge detection operators

Ankit Garg, Anuj Kumar Singh

Summary: This paper introduces the seam diversion-based image retargeting (SDIR) algorithm and analyzes its performance through two experiments. Experiment 1 analyzes the impact of different edge detection operators on the algorithm, while Experiment 2 evaluates the performance based on the visual quality and quality of importance maps. The results show that the SDIR algorithm performs well in minimizing structural deformations on prominent objects.

MULTIMEDIA TOOLS AND APPLICATIONS (2023)

Add to Collection

Article Mathematics, Applied

HORSESHOE PRIORS FOR EDGE-PRESERVING LINEAR BAYESIAN INVERSION

Felipe Uribe, Yiqiu Dong, Per Christian Hansen

Summary: This paper investigates the application of the shrinkage horseshoe prior in edge-preserving settings and introduces its formulation. A Gibbs sampling framework is used to solve the hierarchical formulation of the Bayesian inverse problem, with one conditional distribution being high-dimensional Gaussian and the rest derived in closed form using a scale mixture representation of the heavy-tailed hyperpriors. Applications in imaging science demonstrate that our computational procedure is able to compute sharp edge-preserving posterior point estimates with reduced uncertainty.

SIAM JOURNAL ON SCIENTIFIC COMPUTING (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Similarity activation map for co-salient object detection

Yu Wang, Shuxiao Li

Summary: Co-salient object detection aims to detect common objects in a group of relevant images by leveraging spatial similarity, highlighting common objects while suppressing others and the background. This paper introduces a novel and effective Similarity Activation Module (SAM), as well as Edge Extraction Module (EEM) and Feature Fusion Module (FFM) that can be easily applied to existing methods.

PATTERN RECOGNITION LETTERS (2022)

Add to Collection

Article Engineering, Electrical & Electronic

Empowering ultrasound image filtering precision by reducing speckles and preserving edge cues

Ismail Hababeh, Lina R. Hammad, Mohammad I. Daoud, Mahasen S. Al-Najar

Summary: An efficient adaptive ultrasound image filter is proposed to reduce ultrasound speckle while maintaining the edge cues in the image. The filtering method adapts different window sizes and bandwidths based on the edge map and image radiation. Its performance is compared with eight existing image filtering methods, and the results show its potential to achieve effective speckle suppression in ultrasound images.

INTERNATIONAL JOURNAL OF IMAGING SYSTEMS AND TECHNOLOGY (2023)

Add to Collection

Article Computer Science, Software Engineering

Structure-preserving image smoothing via contrastive learning

Dingkun Zhu, Weiming Wang, Xue Xue, Haoran Xie, Gary Cheng, Fu Lee Wang

Summary: Image smoothing is an important operation to enhance low-frequency structural parts and reduce noise and high-frequency textures in an image. This paper proposes a structure-preserving image smoothing network, which combines paired unsmoothed/smoothed images and meaningful edge information to improve performance. The network is trained using contrastive learning on the extended BSD500 dataset and utilizes an edge-aware total variation loss function for structure preservation. Experimental results show that our network outperforms state-of-the-art image smoothing methods in terms of SSIM and PSNR.

VISUAL COMPUTER (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Position Fusing and Refining for Clear Salient Object Detection

Xing Zhao, Haoran Liang, Ronghua Liang

Summary: Multilevel feature fusion is crucial in salient object detection. This article proposes a global position embedding attention (GPEA) module, an object refine attention (ORA) module, and a pixel value (PV) loss to improve the effectiveness of salient object detection.

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (2022)

Add to Collection

Article Optics

Edge-preserving light-field image super-resolution via feature affine transformation network

Rong Zou, Wenjie Dai, Shenghe Bai, Senlin Mu

Summary: In this study, a novel method for super-resolution of LF images is proposed, which uses both the macro pixel image (MacPI) and the sub-aperture image (SAI) of the LF. The edge map is introduced into the condition network to maintain the edge features of the image. A feature affine-transformation module is designed to guide the features extracted from the SAIs. The super-resolution LF image is recovered using an up-sampling network.

OPTICS AND LASERS IN ENGINEERING (2024)

Add to Collection

Article Engineering, Electrical & Electronic

SPRINT: Spectra Preserving Radiance Image Fusion Technique using holistic deep edge spatial attention and Minnaert guided Bayesian probabilistic model

Indranil Misra, Mukesh Kumar Rohil, S. Manthira Moorthi, Debajyoti Dhar

Summary: Remote Sensing Image Fusion is achieved by merging panchromatic and multispectral images to create a high resolution multispectral image. Existing techniques lack in preserving the spectral characteristics of the multispectral image in the fused output. The proposed Spectra Preserving Radiance Image Fusion Technique (SPRINT) addresses this limitation through holistic deep edges and terrain guidance to maintain spectral fidelity. Evaluation of SPRINT using various datasets shows superior fusion performance compared to state-of-the-art methods, with close agreement in vegetation index and surface reflectance measurements.

SIGNAL PROCESSING-IMAGE COMMUNICATION (2023)

Add to Collection

Article Engineering, Electrical & Electronic

A Privacy-Preserving Federated Learning Method for Probabilistic Community-Level Behind-the-Meter Solar Generation Disaggregation

Jun Lin, Jin Ma, Jianguo Zhu

Summary: Accurate estimation of residential solar photovoltaic generation is crucial for power distribution and demand response programs. A novel method using a federated learning-based Bayesian neural network (FL-BNN) is proposed to disaggregate BTM solar generation at the community level, preserving utility privacy. The effectiveness of the method is validated on a publicly available dataset.

IEEE TRANSACTIONS ON SMART GRID (2022)

Add to Collection

Article Genetics & Heredity

Probabilistic edge inference of gene networks with markov random field-based bayesian learning

Yu-Jyun Huang, Rajarshi Mukherjee, Chuhsing Kate Hsiao

Summary: This study proposes a method that combines Bayesian Markov random field with conditional autoregressive model to simultaneously handle the uncertainty of edge existence and the relative strength of edges in gene regulatory networks. The proposed approach can also prioritize the importance of edges based on their strength. Performance evaluation using simulations and a glioblastoma cancer study shows that the proposed method performs stably and may provide novel insights into biological structures.

FRONTIERS IN GENETICS (2022)

Add to Collection

No Data Available

Article Engineering, Electrical & Electronic

Image tone mapping based on clustering and human visual system models

Xueyu Han, Ishtiaq Rasool Khan, Susanto Rahardja

Summary: This paper proposes a clustering-based TMO method by embedding human visual system models to adapt to different HDR scenes. The method reduces computational complexity using a hierarchical scheme for clustering and enhances local contrast by superimposing details and controlling color saturation by limiting the adaptive saturation parameter. Experimental results show that the proposed method achieves improvements in generating high quality tone-mapped images compared to competing methods.

SIGNAL PROCESSING-IMAGE COMMUNICATION (2024)

Add to Collection

Article Engineering, Electrical & Electronic

YOLO-PAI: Real-time handheld call behavior detection algorithm and embedded application

Zuopeng Zhao, Tianci Zheng, Kai Hao, Junjie Xu, Shuya Cui, Xiaofeng Liu, Guangming Zhao, Jie Zhou, Chen He

Summary: The research team developed a handheld phone detection network called YOLO-PAI, which successfully achieved real-time detection and underwent testing under various conditions. Experimental results show that YOLO-PAI reduces network structure parameters and computational costs while maintaining accuracy, outperforming other popular networks in terms of speed and accuracy.

SIGNAL PROCESSING-IMAGE COMMUNICATION (2024)

Add to Collection

Article Engineering, Electrical & Electronic

ClGanNet: A novel method for maize leaf disease identification using ClGan and deep CNN

Vivek Sharma, Ashish Kumar Tripathi, Purva Daga, M. Nidhi, Himanshu Mittal

Summary: In this study, a novel ClGan method is proposed for automated plant disease detection. The method reduces the number of parameters and addresses the issues of vanishing gradients, training instability, and non-convergence by using an encoder-decoder network. Additionally, an improved loss function is introduced to stabilize the learning process and optimize weights effectively. Furthermore, a new plant leaf classification method called ClGanNet is introduced, achieving 99.97% training accuracy and 99.04% testing accuracy using the least number of parameters.

SIGNAL PROCESSING-IMAGE COMMUNICATION (2024)

Add to Collection

Article Engineering, Electrical & Electronic

Individual tooth segmentation in human teeth images using pseudo edge-region obtained by deep neural networks

Seongeun Kim, Chang-Ock Lee

Summary: This article introduces a method for segmenting individual teeth in human teeth images by using deep neural networks to obtain pseudo edge-regions and applying active contour models for segmentation.

SIGNAL PROCESSING-IMAGE COMMUNICATION (2024)

Add to Collection

© Peeref 2019-2024. All rights reserved.