• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2019년도 하계학술대회
• /
• pp.98-101
• /
• 2019

At present, deep convolutional neural networks have made a very important contribution in single-image super-resolution. Through the learning of the neural networks, the features of input images are transformed and combined to establish a nonlinear mapping of low-resolution images to high-resolution images. Some previous methods are difficult to train and take up a lot of memory. In this paper, we proposed a simple and compact deep recursive residual network learning the features for single image super resolution. Global residual learning and local residual learning are used to reduce the problems of training deep neural networks. And the recursive structure controls the number of parameters to save memory. Experimental results show that the proposed method improved image qualities that occur in previous methods.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2018년도 하계학술대회
• /
• pp.63-66
• /
• 2018

Recently, a convolutional neural network (CNN) models at single image super-resolution have been very successful. Residual learning improves training stability and network performance in CNN. In this paper, we compare four convolutional neural network models for super-resolution (SR) to learn nonlinear mapping from low-resolution (LR) input image to high-resolution (HR) target image. Four models include general CNN model, global residual learning CNN model, local residual learning CNN model, and the CNN model with global and local residual learning. Experiment results show that the results are greatly affected by how skip connections are connected at the basic CNN network, and network trained with only global residual learning generates highest performance among four models at objective and subjective evaluations.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2018년도 추계학술대회
• /
• pp.73-76
• /
• 2018

A Gaussian noise is caused by surrounding environment or channel interference when transmitting image. The noise reduces not only image quality degradation but also high-level image processing performance. The Non-Local Means (NLM) filter finds similarity in the neighboring sets of pixels to remove noise and assigns weights according to similarity. The weighted average is calculated based on the weight. The NLM filter method shows low noise cancellation performance and high complexity in the process of finding the similarity using weight allocation and neighbor set. In order to solve these problems, we propose an algorithm that shows an excellent noise reduction performance by using Summed Square Image (SSI) to reduce the complexity and applying the weighting function based on a cosine Gaussian kernel function. Experimental results demonstrate the effectiveness of the proposed algorithm.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2021년도 추계학술대회
• /
• pp.41-44
• /
• 2021

In this paper, we propose an efficient reference-based compression artifact reduction network for 360-degree images in an equi-rectangular projection (ERP) domain. In our insight, conventional image restoration methods cannot be applied straightforwardly to 360-degree images due to the spherical distortion. To address this problem, we propose an adaptive disparity estimator using a deformable convolution to exploit correlation among 360-degree images. With the help of the proposed convolution, the disparity estimator establishes the spatial correspondence successfully between the ERPs and extract matched textures to be used for image restoration. The experimental results demonstrate that the proposed algorithm provides reliable high-quality textures from the reference and improves the quality of the restored image as compared to the state-of-the-art single image restoration methods.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2007년도 동계학술대회
• /
• pp.68-71
• /
• 2007

자연스러운 가상현실 제작을 위해서는 정확한 카메라 보정(camera calibration) 과정이 필수적인 선결 조건으로 요구된다. 그러나 기존의 영상처리에 의한 카메라 보정 방식은 특징점 추출에서의 에러 발생과 여러 장의 영상을 촬영해야 하는 등의 단점으로 줌렌즈 카메라 보정에는 사용되기 힘들었다. 본 논문에서는 카메라보정 객체의 모델에 기반하여 카메라 파라미터를 최적화하는 방법으로 카메라 보정을 구현하였다 최적화 방법으로는 경사기반 방식에 비해 국부최적점에 강인한 것으로 알려진 유전자알고리즘(genetic algorithm)을 사용하였다. 카메라 보정 객체에 낮은 공간주파수성분을 보강하고, 목적함수에 영상의 밝기 정보를 포함하며, 유전자알고리즘을 사용함으로써 초기치가 최적점에서 멀리 떨어져있는 경우에도 수렴이 가능함을 실험적으로 확인하였다.

• 방송공학회논문지
• /
• 제19권5호
• /
• pp.656-664
• /
• 2014

안개 영상은 영상의 대비가 밝은 영역에 치우쳐 있기 때문에 영상의 정보를 전달하기 어렵다. 이러한 이유로 안개 제거 알고리즘이 연구되고 있다. 일반적으로 안개가 포함되기 전 상태의 영상을 획득하는 것이 어렵기 때문에 알고리즘의 성능을 평가하기 위해 결과 영상을 정성적으로 분석하였다. 본 논문에서는 영상의 변위 정보를 이용하여 안개 영상을 생성함으로써 정량적으로 오차를 비교하는 방법을 제안한다. 또한 이때 은닉 랜덤 마코프 모델(HRMF)에 기반한 기대값 최대화(EM) 알고리즘을 이용하여 블록 결함을 제거하였다. 다양한 합성영상 및 자연영상에 대하여 결과를 비교함으로써 제안한 알고리즘의 성능을 확인하였다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2009년도 IWAIT
• /
• pp.727-731
• /
• 2009

In this paper, we propose a method to generate a free viewpoint image using multi-viewpoint images which are taken by cameras arranged circularly. In past times, we have proposed the method to generate a free viewpoint image based on Ray-Space method. However, with that method, we can not generate a walk-through view seen from a virtual viewpoint among objects. The method we propose in this paper realizes the generation of such view. Our method gets information of the positions of objects using shape from silhouette method at first, and selects appropriate cameras which acquired rays needed for generating a virtual image. A free viewpoint image can be generated by collecting rays which pass over the focal point of a virtual camera. However, when the requested ray is not available, it is necessary to interpolate it from neighboring rays. Therefore, we estimate the depth of the objects from a virtual camera and interpolate ray information to generate the image. In the experiments with the virtual sequences which were captured at every 6 degrees, we set the virtual camera at user's choice and generated the image from that viewpoint successfully.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2009년도 IWAIT
• /
• pp.59-62
• /
• 2009

In general, as a dynamic range of digital still camera is narrower than a real scene‘s, it is hard to represent the shadow region of scene. Thus, multi-scaled retinex algorithm is used to improve detail and local contrast of the shadow region in an image by dividing the image by its local average images through Gaussian filtering. However, if the chromatic distribution of the original image is not uniform and dominated by a certain chromaticity, the chromaticity of the local average image depends on the dominant chromaticity of original image, thereby the colors of the resulting image are shifted to a complement color to the dominant chromaticity. In this paper, a modified multi-scaled retinex method to reduce the influence of the dominant chromaticity is proposed. In multi-scaled retinex process, the local average images obtained by Gaussian filtering are divided by the average chromaticity values of the original image in order to reduce the influence of dominant chromaticity. Next, the chromaticity of illuminant is estimated in highlight region and the local average images are corrected by the estimated chromaticity of illuminant. In experiment, results show that the proposed method improved the local contrast and detail without color distortion.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 1999년도 KOBA 방송기술 워크샵 KOBA Broadcasting Technology Workshop
• /
• pp.133-140
• /
• 1999

In this paper, we propose a method of image-based three-dimensional modeling for realistic facial expression. In the proposed method, real human facial images are used to deform a generic three-dimensional mesh model and the deformed model is animated to generate facial expression animation. First, we take several pictures of the same person from several view angles. Then we project a three-dimensional face model onto the plane of each facial image and match the projected model with each image. The results are combined to generate a deformed three-dimensional model. We use the feature-based image metamorphosis to match the projected models with images. We then create a synthetic image from the two-dimensional images of a specific person's face. This synthetic image is texture-mapped to the cylindrical projection of the three-dimensional model. We also propose a muscle-based animation technique to generate realistic facial expression animations. This method facilitates the control of the animation. lastly, we show the animation results of the six represenative facial expressions.

• 방송공학회논문지
• /
• 제24권6호
• /
• pp.1096-1112
• /
• 2019

현재 360 VR 이미지를 만들어주는 카메라들은 상당히 고가이기에 사람들이 손쉽게 사용할 순 없는 상황이다. 이를 해결하기 위해 휴대 전화의 카메라를 이용해 100여 장의 사진을 360° 촬영을 한 후 Image stitching으로 360 VR 영상을 얻고자 한다. 기존의 장비는 한 번에 360℃ 촬영으로 VR 영상을 만들어내는 반면 휴대 전화를 이용하여 촬영할 경우 영상마다 시차가 생기게 된다. 이로 인해 움직이는 물체가 있는 경우 물체가 여러 장의 영상에서 나타나는 원하지 않는 상황이 생기게 되고 Seam이 물체를 관통하여 부자연스러운 결과 영상을 얻게 된다. 본 논문에서는 시각적으로 두드러지는 물체를 판별할 수 있는 Saliency map을 이용한 Seam finder 알고리즘을 통해 개선된 결과 영상을 얻을 수 있음을 확인했다.