• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2018년도 추계학술대회
• /
• pp.59-61
• /
• 2018

Highly compressed images typically not only have low resolution, but are also affected by compression artifacts. Performing image super-resolution (SR) directly on highly compressed image would simultaneously magnify the blocking artifacts. In this paper, a SR method based on deep learning is proposed. The method is an end-to-end trainable deep convolutional neural network which performs SR on compressed images so as to reduce compression artifacts and improve image resolution. The proposed network is divided into compression artifacts removal (CAR) part and SR reconstruction part, and the network is trained by three-step training method to optimize training procedure. Experiments on JPEG compressed images with quality factors of 10, 20, and 30 demonstrate the effectiveness of the proposed method on commonly used test images and image sets.

• 융합신호처리학회논문지
• /
• 제9권1호
• /
• pp.18-23
• /
• 2008

본 논문에서는 SR(Super Resolution) 복원 과정에 있어 사용되는 입력 후보 영상 중 적합한 입력 영상을 자동 선택하는 알고리즘을 제안함으로써 복원된 고해상도 영상의 질을 개선하고자 한다. SR 복원과정에서 이상적인 결과 영상을 얻기 위해서는 입력되는 모든 영상이 유기적으로 잘 정합 되어야 하지만, 실제로는 그렇지 못하다. 이런 이유로 입력 후보군 영상의 정합 적합성이 얼마나 높은가가 단순히 많은 입력 영상의 수보다 고품질의 고해상도 결과 영상을 얻는데 더욱 결정적이라 할 수 있다. 입력 영상의 적합성은 통계 특성 및 정합 특성을 이용하여 평가 가능하다. 그러므로 본 논문에서는 SR 복원과정에 정합 적합성을 자동으로 평가하여 이에 따라 입력 영상을 결정하는 전처리 과정을 제안하고 구조화하였다. 또한 비디오 시퀀스의 모든 입력 영상은 SR 복원과정의 기준 영상이나 저해상도 입력 영상과 같이 사용될 수 있으므로 본 논문에서는 연속적인 비디오 시퀀스를 위한 SR 복원알고리즘을 제안한다. 적합성의 유무는 임계값(Threshold Value)에 의해 결정되며, 이 임계값은 기준 영상과의 움직임 추정에서 그 보상 값의 오류 값 중 최대치(MMCE, Maximum Motion Compensation Error)로 결정된다. 만약 저해상도 입력 영상의 보상 오류 값의 범위가 0과 MMCE사이(0 < MCE < MMCE )값이라면 그 범위 안의 입력 후보 영상은 SR 복원과정에 사용되며 범위 밖의 후보영상은 제외된다. 최적의 저해상도 기준(ORLR, Optimal Reference Low Resolution)영상은 선택된 저해상도 입력(SLRI, Selected LR Input)영상들과 각각의 저해상도 기준 입력(RLRI, Reference Low Resolution Input)영상들의 비교를 통해 결정된다. 본 논문에서는 이와 같은 과정에 의해 결정된 저해상도의 최적 기준영상과 선택영상을 'Hardie' 보간법을 사용하여 고해상도 영상을 만들어 내는 것으로 사용자의 조정이 없이도 SR 복원영상의 질적 향상을 가져올 것이라 기대된다.

• 전자공학회논문지
• /
• 제50권4호
• /
• pp.137-143
• /
• 2013

고해상도 위성영상은 기상관측, 지형관측, 원격탐사, 군사시설감시, 문화재보호 등 많은 분야에서 이용된다. 위성영상은 동일한 위성영상 시스템에서 획득한 영상이라 할지라도 하드웨어(광학장치, 위성의 운용고도, 영상 센서 등)의 조건에 따라서 해상도가 저하된 영상들이 발생한다. 따라서 위성이 발사된 이후에는 이러한 해상도가 저하된 영상들의 해상도 향상을 위해서 영상시스템의 하드웨어를 변경하는 것은 불가능하므로 위성영상 자체를 이용하여 해상도를 향상시키는 방법이 필요하다. 본 논문에서는 이러한 저해상도 위성영상을 이용하여 해상도를 향상시키는 방법으로 SR(Super Resolution) 알고리즘을 사용하였다. SR 알고리즘은 다수의 저해상도 영상들의 정합을 통해 영상의 해상도를 향상시키는 알고리즘이다. 하지만 위성영상에서는 동일 지역에 대한 여러 장의 영상을 획득하기 어렵다. 따라서 본 논문에서는 이러한 문제점을 해결하기 위해 어파인 변환(Affine Transform)및 투영 변환(Projection Transform)을 적용 후 영상에 대한 기하학적 변화를 보정하여 SR 알고리즘을 수행하였다. 그 결과 SR 알고리즘만 적용한 영상보다 어파인 변환과 투영 변환을 거친 후 SR 알고리즘을 적용한 영상에서 해상도가 확실하게 더 증가되는 것을 확인하였다.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 춘계학술발표대회
• /
• pp.479-481
• /
• 2023

Generative Adversarial Networks (GANs) have facilitated substantial improvement in single-image super-resolution (SR) by enabling the generation of photo-realistic images. However, the high memory requirements of GAN-based SRs (mainly generators) lead to reduced performance and increased energy consumption, making it difficult to implement them onto resource-constricted devices. In this study, we propose an efficient and compressed architecture for the SR-GAN (generator) model using the model compression technique Knowledge Distillation. Our approach involves the transmission of knowledge from a heavy network to a lightweight one, which reduces the storage requirement of the model by 58% with also an increase in their performance. Experimental results on various benchmarks indicate that our proposed compressed model enhances performance with an increase in PSNR, SSIM, and image quality respectively for x4 super-resolution tasks.

• Journal of Multimedia Information System
• /
• 제8권4호
• /
• pp.203-210
• /
• 2021

Deep leaning convolutional neural networks (CNN) have successfully been applied to image super-resolution (SR). Despite their great performances, SR techniques tend to focus on a certain upscale factor when training a particular model. Algorithms for single model multi-scale networks can easily be constructed if images are upscaled prior to input, but sub-pixel convolution upsampling works differently for each scale factor. Recent SR methods employ multi-scale and multi-path learning as a solution. However, this causes unshared parameters and unbalanced parameter distribution across various scale factors. We present a multi-scale single-path upsample module as a solution by exploiting the advantages of sub-pixel convolution and interpolation algorithms. The proposed model employs sub-pixel convolution for the highest scale factor among the learning upscale factors, and then utilize 1-dimension interpolation, compressing the learned features on the channel axis to match the desired output image size. Experiments are performed for the single-path upsample module, and compared to the multi-path upsample module. Based on the experimental results, the proposed algorithm reduces the upsample module's parameters by 24% and presents slightly to better performance compared to the previous algorithm.

• 융합신호처리학회논문지
• /
• 제23권1호
• /
• pp.8-14
• /
• 2022

딥러닝이 발전하면서 초해상화 기술은 단순 보간법(Interpolation)에서 벗어나 딥러닝을 활용해 발전하고 있다. 딥러닝을 사용한 초해상화 기술은 합성곱 신경망(Convolutional Neural Network, CNN) 기반의 연구가 일반적이지만, 최근에는 어텐션(Attention) 메커니즘을 활용한 초해상화 연구가 활발히 진행되고 있다. 본 논문에서는 어텐션 메커니즘 중 하나인 비지역 희소 어텐션(Non-Local Sparse Attention, NLSA)을 활용한 초해상화 성능 향상 방법을 제안한다. 실험을 통해 NLSA를 함께 활용하면 기존 초해상화 신경망 모델인 IMDN, CARN, OISR-LF-s의 성능이 향상되는 것을 확인할 수 있었다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2005년도 Proceedings of ISRS 2005
• /
• pp.197-200
• /
• 2005

Super resolution image reconstruction method refers to image processing algorithms that produce a high resolution(HR) image from observed several low resolution(LR) images of the same scene. This method is proved to be useful in many practical cases where multiple frames of the same scene can be obtained, such as satellite imaging, video surveillance, video enhancement and restoration, digital mosaicking, and medical imaging. In this paper we applied super resolution reconstruction method in spatial domain to video sequences. Test images are adjacently sampled images from continuous video sequences and overlapped for high rate. We constructed the observation model between the HR images and LR images applied by the Maximum A Posteriori(MAP) reconstruction method that is one of the major methods in the super resolution grid construction. Based on this method, we reconstructed high resolution images from low resolution images and compared the results with those from other known interpolation methods.

• 반도체디스플레이기술학회지
• /
• 제19권4호
• /
• pp.71-76
• /
• 2020

In this paper, we propose a super-resolution method that reconstructs compressed low-resolution images into high-resolution images. We propose a CNN model with a small number of parameters, and even if quantization is applied to the proposed model, super-resolution can be implemented without deteriorating the image quality. To further improve the quality of the compressed low-resolution image, a new degradation model was proposed instead of the existing bicubic degradation model. The proposed degradation model is used only in the training process and can be applied by changing only the parameter values to the original CNN model. In the super-resolution image applying the proposed degradation model, visual artifacts caused by image compression were effectively removed. As a result, our proposed method generates higher PSNR values at compressed images and shows better visual quality, compared to conventional CNN-based SR methods.

• 천문학회보
• /
• 제44권2호
• /
• pp.70.4-70.4
• /
• 2019

Image super-resolution (SR) is a technique that enhances the resolution of a low resolution image. In this study, we use three SR models (RCAN, ProSRGAN and Bicubic) for enhancing solar SDO/HMI magnetograms using deep learning. Each model generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). The pixel resolution of HMI is about 0.504 arcsec. Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained three models with HMI images in 2014 and test them with HMI images in 2015. We find that the RCAN model achieves higher quality results than the other two methods in view of both visual aspects and metrics: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is also much better than the conventional bi-cubic interpolation. We apply this model to a full-resolution SDO/HMI image and compare the generated image with the corresponding Hinode NFI magnetogram. As a result, we get a very high correlation (0.92) between the generated SR magnetogram and the Hinode one.

• IEIE Transactions on Smart Processing and Computing
• /
• 제4권1호
• /
• pp.10-15
• /
• 2015

This paper proposes the parallel implementation of a self-similarity based image SR (super-resolution) algorithm using OpenCL. The SR algorithm requires tremendous computations to search for a similar patch. This becomes a bottleneck for the real-time conversion from a FHD image to UHD. Therefore, it is imperative to accelerate the processing speed of SR algorithms. For parallelization, the SR process is divided into several kernels, and memory optimization is performed. In addition, two GPUs are used for further acceleration. The experimental results shows that a GPGPU implementation can speed up over 140 times compared to a single-core CPU. Furthermore, it was confirmed experimentally that utilizing two GPUs can speed up the execution time proportionally, up to 277 times.