Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Joongkyu KimSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works278 from public data
- Automatic Contrast-Limited Adaptive Histogram Equalization With Dual Gamma Correction319
Since automatic CLAHE adaptively enhances contrast in each block while boosting luminance, it is very effective in enhancing dark images and daylight ones with strong dark shadows and outperforms state-of-the-art methods in terms of visual quality and quantitative measures.
- GBDT-MO: Gradient-Boosted Decision Trees for Multiple Outputs236
This article proposes a general method to learn gradient-boosted decision trees for multiple outputs, called GBDT-MO, which achieves outstanding performance in terms of accuracy, training speed, and inference speed.
- 189
- 174
- PIRM Challenge on Perceptual Image Enhancement on Smartphones: Report161
This paper reviews the first challenge on efficient perceptual image enhancement with the focus on deploying deep learning models on smartphones and proposes solutions that significantly improved baseline results defining the state-of-the-art for image enhancement on smartphones.
- 153
- Image deblocking via sparse representation119
Experimental results indicate that the proposed deblocking method is very effective in dealing with the image deblocking problem from compressed images.
- A stroke filter and its application to text localization77
A stroke filter which can detect strokes of texts for robust text localization and remove text candidates which have strong edges but are not text is proposed.
- Retinex-Based Perceptual Contrast Enhancement in Images Using Luminance Adaptation66
Experimental results demonstrate that the proposed retinex-based perceptual contrast enhancement in images using luminance adaptation successfully enhances details in an image while preventing over and under enhancement as well as outperforms state of the arts in terms of various quantitative measurements.
- Low light image enhancement with dual-tree complex wavelet transform63
An effective contrast enhancement method based on dual-tree complex wavelet transform (DT-CWT) which operates on a wide range of imagery without noise amplification and outperforms state-of-the-art ones in terms of contrast enhancement and noise reduction is proposed.
- Head pose estimation using deep neural networks and 3D point clouds61
The proposed method is lightweight and computationally efficient by utilizing a sampled 3D point cloud as input combined with a graph convolutional neural network (GCNN) and achieves outstanding performance in head pose estimation.
- Edge and identity preserving network for face super-resolution58
A novel Edge and Identity Preserving Network for Face Super-Resolution Network, named as EIPNet, which minimizes the distortion by utilizing a lightweight edge block and identity information and defines an identity loss function to preserve identification of super-resolved images.
- 52
- 51
- 49
- NIR to RGB Domain Translation Using Asymmetric Cycle Generative Adversarial Networks44
Experimental results show that the proposed method achieves natural looking NIR colorization results with high generalization ability, and outperforms state-of-the-art ones in realistic colorization and resistance to unregistration.
- Siamese Dense Network for Reflection Removal with Flash and No-Flash Image Pairs41
The proposed siamese dense network (SDN) for reflection removal with flash and no-flash image pairs outperforms state-of-the-art ones in terms of visual quality and quantitative measurements and applies to color/depth image pairs and achieves both color reflection removal and depth filling.
- Accurate text localization in images based on SVM output scores41
A new approach for accurate text localization in images based on SVM (support vector machine) output scores that achieves impressive performance with respect to the accuracy, robustness and efficiency.
- 40
- A new approach for text segmentation using a stroke filter39
This paper attempts to consider the intrinsic characteristics of the text by using the stroke filter and design a new and robust algorithm for text segmentation using a stroke filter based on local region analysis.
- Perceptual Enhancement of Low Light Images Based on Two-Step Noise Suppression38
Experimental results show that the proposed method perceptually enhances contrast in low-light images while successfully minimizing distortions and preserving details, and performs perceptual noise suppression using the JND model.
- FinerPCN: High fidelity point cloud completion network using pointwise convolution37
Experimental results show that FinerPCN generates finer detailed completion results than state-of-the-art methods while successfully keeping the shape of the input.
- Fusionnet: Multispectral Fusion of RGB and NIR Images Using Two Stage Convolutional Neural Networks36
Experimental results show that the proposed method produces natural looking fusion results by successfully recovering colors and textures, and outperforms state-of-the-art fusion methods in terms of visual quality and quantitative measurements.
- Automatic cardiac MRI segmentation and permutation-invariant pathology classification using deep neural networks and point clouds33
The proposed method successfully segments LV, RV, and Myo from cardiac MRI images and achieves comparable results against several existing ones, and the CDN successfully classifies heart diseases based on point clouds and achieves 92% accuracy on the testing dataset.
- Eye detection under varying illumination using the retinex theory33
A novel framework for illumination invariant eye detection under varying lighting conditions using adaptive smoothing based on the retinex theory and the support vector machine (SVM) classification for eye verification is proposed.
- 32
- Text segmentation based on stroke filter32
This paper gives the definition of text, which is realized in the form of stroke filter based on local region analysis, and proposes a novel method of text segmentation based on stroke filter (SF).
- 30
- 30
- 29
- 28
- WaveDH: Wavelet Sub-bands Guided ConvNet for Efficient Image Dehazing27
WaveDH, a novel and compact ConvNet designed to address this efficiency gap in image dehazing, leverages wavelet sub-bands for guided up-and-downsampling and frequency-aware feature refinement and significantly optimizes computational costs.
- 27
- Weakly-supervised temporal attention 3D network for human action recognition26
A weakly-supervised temporal attention 3D network for human action recognition is proposed, called as TA3DNet, to accelerate 3D convolutional neural networks (3D CNNs) by temporally assigning different importance to each frame.
- 26
- 23
- 22
- 20
- 19
- A fast deconvolution-based approach for single-image super-resolution with GPU acceleration19
This paper provides a novel and efficient deconvolution method which combines the gradient consistency in images with the anisotropic regularization which has been used in motion deblurring to produce a directly parallelizable solution suitable for running on GPU by minimizing redundancy in computing.
- Perceptual multi-exposure image fusion with overall image quality index and local saturation19
Experimental results show that the proposed multi-exposure image fusion method based on human visual perception with overall image quality index (OIQ) and local saturation can successfully enhance image quality and achieve realistic rendering of natural scenes.
- 18
- 18
- RegiNet: Gradient guided multispectral image registration using convolutional neural networks18
Experimental results demonstrate that the proposed method successfully produces registration results as well as outperforms state-of-the-art ones in terms of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).
- Perceptually Optimized Enhancement of Contrast and Color in Images18
A perceptual generalized equalization model is used for the optimization of both color and contrast based on color constancy and contrast enhancement, i.e. base image and its JND map to produce the enhanced image highly correlated with the human visual perception.
- Visual comfort enhancement in stereoscopic 3D images using saliency-adaptive nonlinear disparity mapping18
Visual comfort enhancement in S3D images is proposed using saliency-adaptive nonlinear disparity mapping with visual sensitivity to maintain the disparity-based saliency influence and non linear disparity mapping based on a sigmoid function to minimize disparity distortions.
- Deep feature embedding learning for person re-identification based on lifted structured loss17
This paper adopts lifted structured loss for deep neural networks that makes the network learn better feature embedding by minimizing intra- class variation and maximizing inter-class variation.
- 17
- HDVC: Deep Video Compression With Hyperprior-Based Entropy Coding16
This paper introduces hyperprior-based entropy coding into motion compression and optimize motion vector estimation (i.e. optical flow estimation) using window attention and fast residual channel attention to improve compression efficiency and reduce computational complexity while maintaining visual quality.
- Multi-Spectral Fusion and Denoising of Color and Near-Infrared Images Using Multi-Scale Wavelet Analysis16
The proposed method successfully performed the fusion of RGB and NIR images with noise reduction, detail preservation, and color enhancement with the help of a discrepancy model and wavelet scale map.
- 16
- Multispectral Fusion of RGB and NIR Images Using Weighted Least Squares and Convolution Neural Networks15
The proposed multispectral fusion of RGB and NIR images using weighted least squares (WLS) and convolution neural networks (CNNs) achieves good performance in both noise reduction and detail transfer as well as outperforms state-of-the-art methods in terms of visual quality and quantitative measurements.
- Interactive image segmentation via kernel propagation15
A new approach to interactive image segmentation via kernel propagation (KP), called KP Cut, which is based on a learned kernel, and thus can fit the given data better than a predefined kernel.
- WCDANN: A Lightweight CNN Post-Processing Filter for VVC-Based Video Compression14
WCDANN is a convolutional neural network (CNN)-based post-processing filter to enhance the quality of versatile video coding (VVC)-decoded videos without requiring any codec changes.
- 14
- 14
- 13
- Super‐speed up robust features image geometrical registration algorithm13
Experimental results indicate that super-SURF image geometrical registration algorithm is of higher matching accuracy with less running time than SURF algorithm.
- Boundary-preserving stereo matching with certain region detection and adaptive disparity adjustment13
Experimental results demonstrate that the proposed boundary-preserving stereo matching using certain region detection and adaptive disparity adjustment produces high-quality disparity maps by successfully preserving disparity discontinuities along the object boundaries.
- Learning to Rank with Ensemble Ranking SVM13
This paper proposes a novel learning to rank method using Ensemble Ranking SVM, which remarkably improves the efficiency of the model training as well as achieves high ranking accuracy.
- 12
- Part-Level Convolutional Neural Networks for Pedestrian Detection Using Saliency and Boundary Box Alignment12
Part-level convolutional neural networks for pedestrian detection using saliency and boundary box (BB) alignment are proposed and demonstrate that the proposed method remarkably improves accuracy in pedestrian detection and outperforms the existing state-of-the-art techniques.
- Fast Fourier Transform Networks for Object Tracking Based on Correlation Filter12
Experimental results demonstrate that FFTNet outperforms state-of-the-arts in both tracking accuracy and computational efficiency.
- 12
- 12
- 12
- High dynamic range imaging on mobile devices using fusion of multiexposure images12
A simple but effective method to achieve high dynamic range (HDR) rendering results from three multiexposure images comprising under-, normal-, and over-exposure is proposed.
- 12
- 11
- Adaptive tone mapping for display enhancement under ambient light using constrained optimization11
Experimental results demonstrate that the proposed adaptive tone mapping for display enhancement under ambient light significantly enhances the brightness, contrast, details, and colors of displayed images and outperforms other state-of-the-art methods under ambientLight.
- Interactive image retrieval using constraints11
Experimental results demonstrate that the proposed framework remarkably improves the image retrieval performance by ACSP-based constraint propagation in terms of both effectiveness and efficiency.
- 11
- Novel Bayesian deringing method in image interpolation and compression using a SGLI prior11
The proposed deringing method can effectively preserve the significant discontinuities such as textures of objects as well as the strong edge components in images while reducing the ringing artifacts.
- 11
- RTNN: A Neural Network-Based In-Loop Filter in VVC Using Resblock and Transformer10
An NN-based in-loop filter in VVC using Resblock and Transformer, named RTNN, to suppress complicated compression artifacts in VVC is proposed, which combines Resblock and Transformer to extract deep features and capture local and non-local correlation between features.
- NNNet: New Normal Guided Depth Completion From Sparse LiDAR Data and Single Color Image10
Experimental results show that the new normal map provides effective constraints for sparse depth completion and an end-to-end network NNNet, which achieves 724.14 in terms of RMSE and outperforms most of the current state-of-the-art methods.
- 10
- Single Image Depth Estimation Using Edge Extraction Network and Dark Channel Prior10
An edge extraction network based on generative adversarial networks (GANs) to select valid depth edges from a number of edges in an image and distinguish real and fake on the generated depth image using a discriminator, and enhances the performance of the generator.
- A Video Summarization Method for Basketball Game10
Experimental results indicate that the proposed summarization method can summarize basketball video with reasonable accuracy.
- Color Recovery from Multi-Spectral NIR Images Using Gray Information9
The proposed approach is based on multi-spectral NIR images to estimate a conversion matrix for NIR to RGB color conversion and a corresponding gray image is used as a complementary channel to estimate the conversion matrix.
- Disparity-based just-noticeable-difference model for perceptual stereoscopic video coding using depth of focus blur effect9
This work proposes a disparity-based just-noticeable-difference model (DJND) to save bit-rate and improve visual comfort in S3D videos, and combines the DOF blur effect with conventional JND models in the pixel domain into DJND.
- Perceptual rate distortion optimisation for video coding using free‐energy principle9
A free-energy based just-notable-difference model is introduced into the RDO process to adaptively adjust the Lagrange multiplier and quantisation parameter values and results show that the proposed method conceals more distortions in disorderly regions and achieves 8.46% bit-rate reduction over the JM reference software at the same perceptual quality level.
- 9
- Adaptive Constraint Propagation for Semi-Supervised Kernel Matrix Learning9
Experimental results demonstrate that ACP outperforms state-of-the-art SS-KML methods such as PCP and KP in terms of both effectiveness and efficiency.
- 2D to 3D conversion with motion-type adaptive depth estimation9
Experimental results demonstrate that the proposed 2D to 3D conversion with motion-type adaptive depth estimation is very effective in generating accurate depth maps and providing realistic 3D effects.
- Lightweight Generative Network for Image Inpainting Using Feature Contrast Enhancement8
A lightweight generative network for image inpainting using feature contrast enhancement based on dilated convolution and channel attention is proposed and outperforms state-of-the-art ones in terms of visual quality and quantitative measurements.
- WCDGAN: Weakly Connected Dense Generative Adversarial Network for Artifact Removal of Highly Compressed Images8
Experimental results show that WCDGAN successfully removes compression artifacts and produces sharp edges, clear textures and vivid colors even in highly compressed images, and outperforms state-of-the-art methods for compression artifact removal in terms of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).
- Subband Adaptive Image Deblocking Using Wavelet Based Convolutional Neural Networks8
The proposed method achieves comparable state-of-the-art performance on compression artifact removal in terms of both visual quality and quantitative measurements.
- 8
- Semi-supervised kernel matrix learning using adaptive constraint-based seed propagation8
This paper utilizes adaptive constraints to effectively maintain the inherent coherence of samples and successfully propagate constraint information into all samples of semi-supervised kernel matrix learning (SS-KML) using adaptive constraint-based seed propagation (ACSP).
- Multiple Connected Residual Network for Image Enhancement on Smartphones8
Experimental results demonstrate that the proposed method outperforms the state-of-the-art approaches in terms of the perceptual quality and runtime.
- Intensity Guided Depth Upsampling Using Edge Sparsity and Super-Weighted $L_0$ Gradient Minimization8
Experiments on Middlebury and real-world scene datasets verify that the proposed method produces edge preserving depth upsampling results and outperforms state-of-the-arts in terms of both visual quality and quantitative measurements.
- 8
- 8
- Complex Form of Local Orientation Plane for Visual Object Tracking8
A complex form of local orientation plane (Comp-LOP) is proposed, which effectively considers the spatiotemporal relationship between the target and its surrounding region in a correlation filter framework by the complex form, which successfully deals with the heavy occlusion problem.
- 8
- Intensity-guided edge-preserving depth upsampling through weighted L 0 gradient minimization8
Experimental results demonstrate that the proposed intensity-guided edge-preserving depth upsampling method through weighted L0 gradient minimization outperforms state-of-the-art ones in terms of accuracy.
- 8
- Depth-preserving DIBR based on disparity map over T-DMB8
Experimental results show that the proposed DIBR method can create depth-preserving auto-stereoscopic images in mobile broadcasting environments with limited bandwidth.
- Reliable depth-image-based rendering using parameter approximation in mobile devices8
Various experiments show that the proposed DIBR technique can be employed for 3D services based on DIBR over terrestrial-digital multimedia broadcasting (T-DMB) over terrestrial-digital multimedia broadcasting (T-DMB).
- 7
- Joint Disparity Estimation and Pseudo NIR Generation From Cross Spectral Image Pairs7
The proposed network is based on unsupervised learning that consists of one encoder and two decoders, which performs both spectral translation and disparity estimation and generates pseudo NIR images nearly the same as the ground truth even in the regions with large spectral difference.
- 7
- 7
- 7
- 7
- HEVC encoder optimization for HDR video coding based on irregularity concealment effect7
Visual comparison verifies that the proposed HEVC encoder optimization for HEVC Main 10 Profile-based HDR video coding preserves more structural and edge information than HEVC in reconstructing HDR videos.
- 7
- Visual comfort assessment in stereoscopic 3D images using salient object disparity7
Experimental results show that the proposed method substantially improves the VC prediction performance in S3D images compared with conventional prediction methods.
- 7
- 7
- AGNet: Attention Guided Sparse Depth Completion Using Convolutional Neural Networks6
This paper proposes attention guided sparse depth completion using convolutional neural networks, called AGNet, and adopts attention learning to get geometric cues for depth regression from RGB image and capture multi-scale depth structures.
- 6
- 6
- Scale-Aware Multispectral Fusion of RGB and NIR Images Based on Alternating Guidance6
Experimental results show that the proposed scale-aware multispectral fusion of RGB and NIR images based on alternating guidance achieves good performance in noise reduction, detail transfer and color reproduction, and is superior to the state-of-the-art ones in terms of quantitative measurement and computational efficiency.
- DCM-CNN: Densely Connected Multiloss Convolutional Neural Networks for Light Field View Synthesis6
This paper builds DCM-CNN for view synthesis based on the feature reuse, thereby leading to low computational complexity and presents a multiloss function that consists of pixel loss, feature loss and edge loss to produce accurate edges and textures in the synthesized views.
- Adaptive perceptual quantizer for high dynamic range video compression6
Adaptive transfer function based on PQ for HDR video compression, called adaptive PQ is proposed, which adaptively maps luminance to luma according to HDR contents and achieves better perceptual uniformity in the luminance range than PQ.
- Visual saliency estimation using constraints6
Experimental results demonstrate that kernel learning and seed propagation are capable of learning distinctive saliency from images.
- 6
- 6
- Automatic image segmentation using constraint learning and propagation6
This paper employs kernel propagation (KP) based on kernel learning to effectively learn the constraints generated by initial seeds for image segmentation, and designs a novel segmentation method to achieve high performance.
- Bayesian image denoising using two complementary discontinuity measures6
The experimental results show that the proposed Bayesian denoising framework not only achieves a high peak signal to noise ratio (PSNR) gain from noisy images but also reduces noise effectively while preserving edge components.
- Spatial-gradient-local-inhomogeneity: an efficient image-denoising prior6
Comparative results indicate that the proposed SGLI prior is very effective in dealing with the image denoising problem from corrupted images.
- A New Approach for Bayesian Denoising in Images Using an Object Homogeneity Prior6
The proposed method applies a Bayesian framework for efficient denoising of images corrupted with noise which can, respectively, be used as a backbone to effectively reducing much more complex noise.
- 6
- Deep Sparse Depth Completion Using Multi-Scale Residuals and Channel Shuffle5
This paper proposes deep sparse depth completion using multi-scale residuals and channel shuffle, named ReCSNet, a dual-branch network based on a U-shaped architecture that achieves outstanding performance on the KITTI depth completion validation dataset with an improvement of 16mm in the root mean square error (RMSE) metric.
- Face Reflection Removal Network Using Multispectral Fusion of RGB and NIR Images5
A face reflection removal network using multispectral fusion of color (RGB) and near infrared (NIR) images, called FRRN, which outperforms state-of-the-art methods for reflection removal in terms of visual quality and quantitative measurements.
- Low Complexity In-Loop Filter for VVC Based on Convolution and Transformer5
ConvTransNet enables the NN-based in-loop filter to reduce compression artifacts and enhance visual quality in an image and achieves outstanding performance in terms of BD-rate and complexity.
- 5
- 5
- Quality Enhancement of Compressed 360-Degree Videos Using Viewport-based Deep Neural Networks5
This article proposes a quality enhancement of compressed 360-degree videos using viewport-based deep neural networks, named V-DNN, which consistently outperforms MFQE-1.0, MFQe-2.
- 5
- Guided filtering based data fusion for light field depth estimation with L0 gradient minimization5
Experimental results on both synthetic and real light field datasets show that the proposed method achieves clearer edge and less error in depth than state-of-the-arts.
- 5
- 5
- 5
- 5
- Curvature preserving image super-resolution with gradient-consistency-anisotropic-regularization prior5
This paper aims to achieve artifact-free SR reconstruction from an input low resolution (LR) image using adaptive de-convolution and curvature refinement and proposes a curvature preserving image SR method based on a gradient-consistency-anisotropic-regularization (GCAR) prior.
- 5
- A Novel Approach for Bayesian Image Denoising Using a SGLI Prior5
This paper provides an effective prior for image denoising in a Bayesian framework that combines two well-known discontinuity measures used in illumination normalization methods and achieves impressive performance with respect to noise reduction.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- Motion segmentation using Markov random field model for accurate moving object segmentation4
A new motion segmentation algorithm for accurate moving object segmentation based on Markov random field (MRF) model based on Bayes rule that achieves good performance with respect to the segmentation accuracy.
- A Novel Approach for Key Caption Detection in Golf Videos Using Color Patterns4
This paper provides a novel method of detecting key captions containing player information in golf videos using the color pattern of captions and its repetition property to determine the key caption.
- Automatic Superimposed Text Localization from Video Using Temporal Information4
This paper presents the new and fast superimposed text localization method in video segments that has minimized the candidate region of localizing superimposed texts by using the difference between consecutive frames.
- 3
- 3
- Deep Sparse Depth Completion Using Multi-Affinity Matrix3
The proposed network based on multi-affinity matrix achieves state-of-the-art performance in terms of root mean square error (RMSE) and mean absolute error (MAE) and predicts accurate depth values by calculating the weights between neighboring pixels.
- Attention-Aware Linear Depthwise Convolution for Single Image Super-Resolution3
An attention-aware linear depthwise network for single image SR, named ALDNet, that allows CNN-based SR models to preserve useful information for reconstructing a super-resolved image while reducing computational burden and makes full use of depthwise filter interdependencies.
- 3
- 3
- 3
- 3
- 3
- Content adaptive video denoising based on human visual perception3
Extensive experiments on various video sequences demonstrate that the proposed method reconstructs natural-looking results even in highly corrupted images and achieves good performance in terms of both visual quality and quantitative measures.
- Enhancing dictionary-based super-resolution using nonlocal total variation regularization3
Experimental results demonstrate that the proposed method is very effective in enhancing the dictionary-based SR approaches in terms of quantitative performance and visual quality.
- High-dynamic-range image rendering using hybrid tone mapping and automatic k factor decision3
This work proposes a novel HDR compression method that combines global and local tone-mapping operators, and manipulates the HDR compression in both luminance and gradient domains to improve the appearance and naturalness of results.
- A novel approach to motion segmentation based on gamma distribution3
A new motion segmentation method in image sequences based on gamma distribution is provided, which is more suitable than the Gaussian mixture model for the optical flow based motion segmentsation.
- 3
- New Optical Flow Approach for Motion Segmentation Based on Gamma Distribution3
A new and accurate motion segmentation method based on the gamma distribution of the velocity vector's magnitude based on a Markov random field for noise reduction and pixel labeling is proposed.
- 3
- 2
- Frequency-Aware Learned Image Compression Using Channel-Wise Attention and Restormer2
This paper proposes a frequency-aware learned image compression network based on channel-wise attention and Restormer, and designs a frequency-aware network based on the discrete wavelet transform in a hyperprior network to learn both spatial and frequency features.
- 2
- MRFGNet: Multiscale Reference Frame Generation Network for VVC Inter-Coding2
A multiscale reference frame generation network for VVC inter-coding, named MRFGNet, which adopts the high-performance operation point (HOP) network, which has been released by JVET, as preprocessing for frame enhancement and estimates optical flow at multiple scales with the same dimension, thereby leveraging scale-independent bidirectional optical flow prediction.
- 2
- 2
- 2
- 2
- Weighted Chroma Downsampling and Luma-Referenced Chroma Upsampling for HDR/WCG Video Coding2
The experimental results show that the proposed weighted chroma downsampling and luma-referenced chroma upsampling for HDR/WCG video coding reconstructs high quality HDR videos with better image details and less color distortion, and outperforms HDR Anchor in terms of both objective and subjective evaluations.
- Gradient Guided Image Deblocking Using Convolutional Neural Networks2
Guided by the gradient information of the input blocky image, the proposed network successfully preserves textural edges while reducing blocky edges, and thus restores the original clean image from compression degradation, and achieves a significant performance improvement in terms of visual quality and objective measurements.
- RSC-DGS2
Experimental results confirm that the proposed method effectively preserves color/spectral consistency and textures in fusion results while successfully removing noise with high computational efficiency.
- 2
- 2
- Recurrent Convolutions: A Model Compression Point of View2
This paper proposes a simple variant which improves RC networks: the batch normalization layers of an RC module are learned independently (not shared) for different unrolling steps, and provides insights on why this works.
- 2
- Perceptual stereoscopic video coding using disparity just-noticeable-distortion model2
Experimental results demonstrate that the proposed method significantly improves coding efficiency without loss of stereoscopic perceptual quality.
- 2
- 2
- 2
- Perceptual block merging for quadtree-based partitioning in HEVC using disorderly concealment effect2
- Visual saliency estimation through label propagation2
Experimental results demonstrate that the proposed method is very effective in estimating salient regions in images compared with existing state-of-the-art methods.
- 2
- Chrominance just‐noticeable‐distortion model based on human colour perception2
Experimental results demonstrate that the proposed chrominance JND model significantly saves bit rate without loss of perceptual quality in stereo videos.
- 2
- 2D to 3D Conversion in 3DTV Using Depth Map Generation and Virtual View Synthesis2
Experimental results demonstrate that the proposed 2D to 3D conversion is very effective in generating depth maps and providing realistic 3D effects.
- Toward high-quality image communications: inverse problems in image processing2
More research is required to overcome the limited bandwidth of current communications systems and achieve high-quality image reconstruction in real applications and discuss practical considerations and possible solutions to implement these two technologies in real mobile applications.
- Specular Highlight Removal Using Reflection Component Separation and Joint Bilateral Filtering2
This work refine the separated diffuse component using a joint bilateral filter and demonstrates that the proposed method outperforms the state-of-the-art ones for highlight removal.
- Specularity removal based on reflection component separation and joint bilateral filtering2
A novel specularity removal method based on reflection component separation and joint bilateral filtering is proposed that outperforms the state-of-the-art ones in specular removal.
- Illumination Invariant Eye Detection in Facial Images Based on the Retinex Theory2
A new framework for illumination invariant eye detection robust against varying lighting conditions is provided and an adaptive smoothing based on the Retinex theory is used to remove the illumination effects.
- 2
- 1
- Pixelated Face Recovery Using Multi-Step Cycle Asymmetric GAN1
Experimental results show that the proposed method effectively recovers clear facial details from the pixelated faces while successfully recognizing the facial identity, and outperforms state-of-the-art (SOTA) ones for face depixelation in terms of both visual quality and quantitative measurements.
- 1
- Low Operation Point In-Loop Filter for VVC Based on Over-Parameterized Training and Variable Channel Configurations1
The proposed method enhances the Bjøntegaard Delta rate (BD-rate) performance while reducing the complexity of LOP network from 16.83k multiply-accumulate (kMAC)/pixel to 16.81 kMAC/pixel.
- 1
- Dual-Branch Neural Network-Based In-Loop Filter for VVC Intra Coding Using Spatial-Frequency Feature Fusion1
A dual-branch NNLF for VVC intra coding using spatial-frequency feature fusion is proposed and fast Fourier transform (FFT) is adopted to capture global context in a frame, instead of Transformer that has relatively high complexity.
- 1
- 1
- 1
- Latent Edge Guided Depth Super-Resolution Using Attention-Based Hierarchical Multi-Modal Fusion1
This paper proposes a latent edge guided depth SR network using attention-based hierarchical multi-modal fusion, named LEDSRNet, and designs an attention based multi-level residual block as the basic block to extract both shallow and deep features in color and depth images for hierarchical multi-modal fusion.
- 1
- 1
- 1
- 1
- 1
- SuperStyleNet: Deep Image Synthesis with Superpixel Based Style Encoder1
The proposed network achieves high-quality image synthesis by mapping the style codes into semantic labels into vectorized style codes based on graphical analysis, and outperforms state-of-the-art ones in terms of visual quality and quantitative measurements.
- 1
- Perceptually motivated fast coding unit size decision for HEVC intracoding based on visual regularity1
This work proposes perceptually motivated fast CU decision for HEVC intracoding based on visual regularity, which saves 24% computational cost on average as compared to anchor and predicts irregular CUs by exploiting the information of neighboring CUs, which would not introduce too much distortion.
- 1
- 1
- Image-guided depth propagation for 2-D-to-3-D video conversion using superpixel matching and adaptive autoregressive model1
An adaptive image-guided AR model is introduced to minimize matching errors and produce the final depth map by minimizing AR prediction errors and demonstrates that the proposed method successfully performs depth propagation and produces high-quality depth maps for 2-D-to-3-D video conversion.
- Rapid learning-based video stereolization using graphic processing unit acceleration1
Experimental results demonstrate that the proposed rapid learning-based video stereolization using a graphic processing unit (GPU) acceleration is nearly 180 times faster than CPU-based processing and achieves a good performance comparable to the-state-of-the-art ones.
- 1
- 1
- 1
- 1
- 1
- Blocking Artifact Reduction in DIBR Using an Overcomplete 3D Dictionary1
This paper proposes a new deblocking method in 3D images, i.e., video-plus-depth, using an overcomplete 3D dictionary, and generates the3D dictionary from natural and depth images using the k-singular value decomposition (K-SVD) algorithm, and estimates an error threshold to utilize the 3D Dictionary using a compression factor of the compressed images.
- Korean-English bilingual videotext recognition for news headline generation based on a split-merge strategy1
A novel Korean-English bilingual videotext recognition method to overcome the computational complexity as well as achieve comparable recognition accuracy and is successfully employed in generating news headlines.
- Automatic Segmentation of Non-rigid Objects in Image Sequences Using Spatiotemporal Information1
This paper provides an automatic segmentation method of non-rigid objects in image sequences using spatiotemporal information and demonstrates that the performance of the method is quite impressive from the viewpoints of the segmentation accuracy.
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Saving Bits Using Multi-Sensor Collaboration–
The proposed method combines CNN-US and CNN-QE with 3D-HEVC to save bits using multi-sensor collaboration and achieves average 5.9%, 66.8%, and 71.0% BD-rate reductions for sampling factors 1, 2, and 4 on the depth videos of 3D -HEVC test dataset, respectively.
- –
- –
- –
- –
- –
- Power constrained contrast enhancement based on brightness compensated contrast-tone mapping operation–
This work adopts brightness compensation to restore the original perceptual luminance and combine tone mapping with contrast enhancement to improve image quality and implements BCCTMO on an Android tablet with an LCD display using mathematical approximation.
- –
- –
- –
- –
- –
- –
- –
- Joint geodesic depth propagation for two-dimensional-to-three-dimensional video conversion–
A key-frame depth propagation strategy based on a geodesic distance is adopted, which performs block matching between current and reference frames to estimate motion vectors (MVs) and conducts depth compensation from the reference depth map for obtaining the current depth map based on MVs.
- –
- Three-dimensional dictionary-based deblocking for high-quality stereoscopic view generation–
Experimental results demonstrate that the proposed method can effectively reduce annoying blocking artifacts in compressed3-D images, i.e., video-plus-depth, and generate high-quality 3-D stereoscopic views.
- 3D DMB Player and Its Reliable 3D Services in T-DMB Systems–
Experimental results show that the parameter approximation method can create auto-stereoscopic images reliably in the 3D DMB player and how to approximate the rendering parameters by taking mobile display size into consideration is explained.
- –
- Automatic Summarization of Basketball Video Using the Score Information–
A method for content based automatic summarization of basketball game videos using the score information in basketball videos and recognizing the digits on the score caption and analyzing the variation of the score is proposed.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.