Sehwan Ki

About My SELF

  • Samsung Advanced Institute of Technology (SAIT)

  • Ph.D. in Electrical Engineering in KAIST

  • E-mail: dkfzkvk1992@gmai.com

  • [CV] [Google Scholar] [LinkedIn] [Github]

Research Interests

  • Perceptual Video and Image Compression

  • Perceptual Video and Image quality Assessment

  • Deep Learning based Video and Image perceptual quality Enhancement: Super-Resolution, HDR conversion, Coding Artifact Removal, Denoising, Deblur, Dehazing, Frame Up-conversion

  • GPU optimization/programming for deep learning: CUDA programming

  • Low complexity Deep Network (Knowledge distillation, Pruning, Quantization)

PUblications

Learning-based JND-directed HDR Video Preprocessing for Perceptually Lossless Compression with HEVC

Sehwan Ki, Jeonghyeok Do, and Munchurl Kim

IEEE Access, vol.8, pp. 228605-228618, Dec 2020

[PDF] [Github]

A Novel Just-Noticeable-Difference-based Saliency-Channel Attention Residual Network for Full-Reference Image Quality Predictions

Soomin Seo*, Sehwan Ki*, and Munchurl Kim (*equal contribution)

IEEE Transactions on Circuits and Systems for Video Technology, Early Access, Oct. 2020

[PDF]

Learning-based Just-Noticeable-Quantization-Distortion Modeling for Perceptual Video Coding

Sehwan Ki, Sung-Ho Bae, Munchurl Kim, and Hyunsuk Ko

IEEE Transactions on Image Processing, vol. 27, no. 7, pp. 3178-3192, July, 2018

[PDF] [Github]

High-Resolution Image Dehazing with respect to Training Losses and Receptive Field Sizes

Hyeonjun Sim, Sehwan Ki, and Munchurl Kim

Computer Vision and Pattern Recognition Workshops (CVPRW), 2018 IEEE Conference on, Salt Lake CIty, UT, USA, 18 June 2018.

(Oral Presentation, 2nd Rank in Outdoor Track)

[PDF] [Poster]

Fully End-to-End learning based Conditional Boundary Equilibrium GAN with Receptive Field Sizes Enlarged for Single Ultra-High Resolution Image Dehazing

Sehwan Ki, Hyeonjun Sim, and Munchurl Kim

Computer Vision and Pattern Recognition Workshops (CVPRW), 2018 IEEE Conference on, Salt Lake CIty, UT, USA, 18 June 2018.

(4th Rank in Indoor Track)

[PDF] [Poster]

NTIRE 2018 Challenge on Image Dehazing: Methods and Results

Cosmin Ancuti, Codruta O. Ancuti, Radu Timofte, Luc Van Gool, Lei Zhang, Ming-Hsuan Yang [and 59 others, including Sehwan Ki]

Computer Vision and Pattern Recognition Workshops (CVPRW), 2018 IEEE Conference on, Salt Lake CIty, UT, USA, 18 June 2018.

[PDF]

NTIRE 2018 Challenge on Single Image Super-Resolution: Methods and Results

Radu Timofte, Shuhang Gu, Jiqing Wu, Luc Van Gool, Lei Zhang, Ming-Hsuan Yang [and 95 others, including Sehwan Ki]

Computer Vision and Pattern Recognition Workshops (CVPRW), 2018 IEEE Conference on, Salt Lake CIty, UT, USA, 18 June 2018.

[PDF]

Just-noticeable-quantization-distortion based preprocessing for perceptual video coding

Sehwan Ki and Munchurl Kim

IEEE International Conference on Visual Communications and Image Processing (VCIP), St Petersburg, Florida, USA, 10-13 Dec. 2017.

[PDF] [Poster]

Domestic Journal

  • Sehwan Ki, Dae-eun Kim and Munchurl Kim, “Performance Analysis of Super-Resolution based Video Coding for HEVC,” Journal of Broadcast Engineering, vol. 24, no. 2, pp. 306-314, Mar 2019.

  • Dae-eun Kim, Sehwan Ki and Munchurl Kim, “Scalable Video Coding using Super-Resolution based on Convolutional Neural Networks for Video Transmission over Very Narrow-Bandwidth Networks,” Journal of Broadcast Engineering, vol. 24, no. 1, pp. 132-141, Jan 2019.

Domestic Conference

  • Sehwan Ki and Munchurl Kim, “Content-Adaptive Model Update of Convolutional Neural Networks for Super-Resolution”, Korea Institute of Broadcast and Media Engineers Workshop, Nov. 2020.

  • Sehwan Ki and Munchurl Kim, “Temporally adaptive and region- selective signaling of applying multiple neural network models”, Korea Institute of Broadcast and Media Engineers Workshop, Nov. 2020.

  • Sehwan Ki and Munchurl Kim, “A Study on the Convolution Neural Network based on Blind High Dynamic Range Image Quality Assessment,” Korea Institute of Communications and Information Sciences Workshop, Dec. 2018.

  • Sehwan Ki, Jae-seok Choi, Sooye Kim and Munchurl Kim, “Accelerating Deep Learning based Super-resolution algorithm using GPU,” Korea Institute of Broadcast and Media Engineers Workshop, June 2017.

  • Sehwan Ki and Munchurl Kim, “Just Noticeable Quantization Blur model on the DCT complexity feature of the image,” Korea Institute of Broadcast and Media Engineers Workshop, June 2016.

  • Sehwan Ki and Munchurl Kim, “JND based Video Preprocessing Adaptive to Quantization Step sizes for Perceptual Redundancy Reduction,” Korea Institute of Broadcast and Media Engineers Workshop, Sept. 2016.

PAtent and Standard

  • IMAGE PROCESSING METHOD AND APPARATUS USING SELECTION UNIT (Domestic), July. 2018

  • Apparatus and Method for Performing Scalable Video Decoding (Domestic), Mar. 2019 (Pending)

  • Sehwan Ki and Munchurl Kim, “Use cases and requirements for neural network compression for multimedia content description and analysis ,” ISO/IEC JTC1/SC29/WG11/N18731, July 2019, Gothenburg, SE. [PDF]

Awards and Honors

  • NTIRE 2018 Challenge on Single Image Dehazing in conjunction with CVPR 2018, 4th Place Award, June 2018 [PDF]

  • NTIRE 2018 Challenge on Single Image Super-Resolution in conjunction with CVPR 2018, Honorable Mention Award, June 2018 [PDF]

  • Kyungpook National University, Excellence Graduate Honor (3/395) , Feb. 2015

  • 2014 KNU Capstone Design Contest 2nd Place Award, Dec. 2014

Education

  • Ph.D. in Electrical Engineering, KAIST, Daejeon, Korea, Mar. 2017 - Feb. 2021 (GPA: 4.13/4.30)

  • M.S. in Electrical Engineering, KAIST, Daejeon, Korea, Mar. 2015 - Feb. 2017 (GPA: 4.12/4.30)

  • B.S. in Electrical Engineering, Kyungpook, Daegu, Korea, Mar. 2011 - Feb. 2015 (GPA: 4.15/4.30)

Major Skills

  • Programming Languages - C, C++, Matlab, Python, CUDA C, OpenGL/CL, SNPE

  • Deep Learning Framework - Pytorch, Tensorflow

  • Subjective Video and Image Quality Assessment

  • GPU Programming

  • Deep learning for Mobile Devices

Project Experience

  • National Research Foundation of Korea, Research Project on Intelligent and Highly Realistic Visual Processing for Smart Broadcasting Media (Super-Resolution, Dehazing, Deep Learning based image/video quality enhancement), Mar. 2017 – Dec. 2020

  • LG, Low complexity deep Learning based super-resolution. Dec. 2018 - Dec. 2019

  • ETRI, Deep Learning based New Generation Video Coding. Mar. 2016 - Dec. 2019

  • LIG Nexone, Deep Learning based Ultra-narrow bandwidth Video Coding, Super-Resolution. Jan. 2018 - Dec. 2019

  • Huawei Technologies Co. Ltd., Deep Learning based Reconstructed Video Coding, GPU programming. Apr. 2017 - Nov. 2019

  • Huawei Technologies Co. Ltd., Deep Learning based Super-Resolution, GPU programming. Sept. 2016 - Mar. 2017