한국어

연구분야

화질 개선 및 화상 처리

n  Image Deblurring

KEYWORD: image deblurring, point spread function (PSF), PSF estimation, deconvolution --------------------------------------------------------------

 

영상 획득 객체의 움직임이나 카메라의 움직임카메라의 상태 등에 따라 다양한 종류의 영상 흐림 (blur) 발생할  있다이러한 영상 흐림은 영상의 화질을 심각하게 저하시키므로 이를 제거하기 위해 다양한 종류의 영상 복원 기법들이 개발되었다영상 흐림을 제거하는 가장 대표적인 기법인 역컨볼루션 기법은 왜곡된 영상을 원래의 선명한 영상과 point spread function (PSF) 표현되는 흐림 현상의 컨볼루션으로 모델링하고이를 역컨볼루션을 이용하여 영상의 화질을 복원하는 기법이다. 
  PSF
 정확히 알고 있다는 가정하에 원영상을 복원하기 위해 다양한 알고리즘이 제안되었으며전통적인 방법으로는Wiener 필터, Lucy-Richardson 알고리즘 등이 유명하다. Wiener 필터는 노이즈를 알고 있다는 가정하에 영상의 주파수 특성을 이용하여 흐림 현상을 제거 하는inverse 필터이다화질 열화 과정에서 노이즈가 없다면 Wiener 필터는 이상적인 inverse 필터와 일치한다. Lucy-Richardson 알고리즘에서는 화질 열화 과정에서Poisson 노이즈가 발생한다고 가정하고모델링한 수식에 의해 구해진 열화 영상과 주어진 열화된 영상의 likelihood  최대화하도록 알고리즘을 반복적으로 수행하여 흐림을 제거한다하지만 이러한 기술이 실제 환경에 적용되기 위해서는 정확한 PSF 예측이 필요하다. 따라서,  연구실에서는 PSF 예측 기법과 영상의 흐림 제거 기법을 연구하고 있다.

 

Related Papers

[1] B.-D. Choi, S.-W. Jung, and S.-J. Ko, "Motion-blur-free camera system splitting exposure time," IEEE Trans. Comsum. Electron., vol. 54, no. 3, pp. 981-986, Aug. 2008.

[2] S.-W. Jung, B.-D. Choi, and S.-J. Ko, “Iterative PSF estimation and Its Application to Shift Invariant and Variant Blur Reduction,” EURASIP Journal on Advances in Signal Processing, vol. 2009, Article ID 909636, 14 pages, Dec. 2009.

[3] S.-W. Jung, T.-H. Kim, and S.-J. Ko, "A Novel Multiple Image Deblurring Technique Using Fuzzy Projection onto Convex Sets," IEEE Signal Processing Letters, vol. 16, no. 3, pp. 192-195, Mar. 2009.


Related Patent

영상의 모션 블러를 제거하기 위한 장치  방법(Apparatus and method for removing motion blur of image), patent no. 10-0860967-0000, Sep. 2008

  

Simple flowchart of image deblurring

image002.png

[*]S. Cho and S. Lee, "Fast motion deblurring," ACM Trans. Graph., vol. 28, no. 5, pp. 145:1-8, 2009.


Comparison of image deblurring performance based on the accuracy of the PSF estimation

image004.png

[**]O. Whyte, J. Sivic, A. Zisserman, and J. Ponce, "Non-uniform deblurring for shaken images, Proc. IEEE Int. Conf. Comput. Vis. and Pattern Recogn.San Francisco, Califonia, USA, Jun. 2010, pp. 1-8.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Current research topics:

- Image restoration using the multi-exposure time images [1]

- PSF estimation using Modulation transfer function (MTF) for satellite images [2]

- Image restoration using the multi-resolution images [3]

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

 

n  Super Resolution (Image Upscaling)

KEYWORD: super resolution, image upscaling, interpolation, reconstruction-based super resolution -------------------------------------------------

 

최근 TV 모바일 기기와 같은 고해상도 디스플레이 장치의 발전과 함께저해상도 영상을 이용하여 고해상도 영상을 생성하는 초해상도 (Super resolution (SR)) 기술이 주목을 받고 있다현재까지는 크게 interpolation-based SR, example-based SR, multi-frame-based SR, reconstruction-based SR 이렇게 4가지의 방향으로 초해상도 기법이 제안되어 왔다초해상도 기술에 있어 가장 중요한 이슈는 1) jagging, blurring, ringing  같은 노이즈가 최소화된 고해상도 영상을 만드는 것과 2) 저해상도 영상에서의 texture 성분이 고해상도 영상에서도 최대한 유사하게 유지되게 하는 것이다기존의 방식들 , interpolation-based SR 기법과 reconstruction-based SR기법은 이러한 화질  복잡도 관점에서 우수한 성능을 보인다 연구실에서는   가지 방식에 초점을 맞추어 효과적인 초해상도 기법을 연구  개발하고 있다.


image006.png

image008.png

 

Related Papers

[1] J.-W. Han, J.-H. Kim, S.-H. Cheon, J.-O. Kim, and S.-J. Ko, "A Novel Image Interpolation Method Using the Bilateral Filter," IEEE Trans. Consumer Electron., vol. 56, no. 1, pp. 175-181, Feb. 2010.

[2] J.-W. Han, Suryanto, J.-H. Kim, S. Sull, and S.-J. Ko, "New Edge-Adaptive Image Interpolation Using Anisotropic Gaussian Filters," Digital Signal Processing, vol. 23, no. 1, pp. 110-117, Jan. 2013.

[3] J.-H. Kim, S.-J. Lee, M.-Y. Cho, Y.-J. Yoon, and S.-J. Ko, “A Second-Order Derivative-Based Edge Preserving Image Interpolation,” IEEE ICCE 2015, Jan. 2015.

[4] 김종환, 이승준, 고성제, "향상된  엣지 보존 능력의 DFDF 영상 보간법," 대한전자공학회 하계학술대회 논문집, 제주, 대한민국, pp. 1867-1870, Jul. 2013.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------


Current research topics:

- Super resolution for mobile devices with high PPI

- Fast super resolution method for UHDTV [3]

- Edge-directed single image interpolation [1-4]

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

 

n  Contrast Enhancement

KEYWORD: high dynamic range (HDR), dynamic range compression ------------------------------------------------------------------------------------------

 

대조비 개선은 영상에서의 밝고 어두운 부분의 대비를 극대화함으로써 영상 표현을 향상시키는 것이다그간 많은 알고리즘들(gray-level transformation based techniques, histogram processing techniques) 이 제안되어 왔는데대부분의 알고리즘들이 global하게 영상을 변화시키는 방식이었기 때문에 화소 값의 변화가 심한 영역에서의 결과가 좋지 못했다따라서 본 연구실에서는 다양한 영상에 효과적으로 적용될 수 있는대조비 개선 알고리즘들을 개발하고 있다.

 

Related Papers

[1] Y.-J. Yoon, K.-Y. Byun, D.-H. Lee, S.-W. Jung, and S.-J. Ko. "A New Human Perception-Based Over-Exposure Detection Method for Color Images," Sensors, vol. 14, no. 9, pp. 17159-17173, Sep. 2014.

[2] D.-H. Lee, Y.-J. Yoon, S.-J. Kang, and S.-J. Ko, "Correction of the Overexposed Region in Digital Color Image," IEEE Trans. Consum. Electron., vol. 60, no. 2, pp. 173-178, May 2014.

[3] J.-H. Kim, H. Kim, and S.-J. Ko, "New Visualization Method for High Dynamic Range Images in Low Dynamic Range Devices," Optical Engineering, vol. 50, no. 10, pp. 107005-1-7, Oct. 2011.

[4] S.-W. Jung, B.-D. Choi, and S.-J. Ko, "A novel approach integrating noise reduction and contrast/sharpness enhancement," Proc. of 2006 International Conference on Electronics Information and Communication, Ulaanbaatar, Mongolia, Jun. 2006.

 

Example of HDR image compression

image010.pngimage012.png

Example of under/over-exposure image correction

image014.png  image016.png

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Current research topics:

- Under/Over-exposure correction [1],[2]

- HDR(High Dynamic Range) imaging [3]

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

 

n  Digital Image Stabilization (DIS)

KEYWORD: digital image stabilization (DIS), global motion estimation, motion filtering, video stabilization ------------------------------------------

 

영상 안정화 기술은 손떨림이나 바람과 같은 외부 환경으로 인해 카메라가 흔들릴  원치 않는 떨림을 제거하여 선명한 영상을 얻을  있게 해주는 기술이다최근의 영상 안정화는 자이로 센서나 유동 프리즘 같은 부가적인 장치만 사용하는 것이 아니라디지털 영상 처리를 이용한 방법이 추가로 적용되어 회전  3차원 왜곡과 같이 기존의 부가 장치만으로 안정화가 불가능했던 영상 왜곡에 대해 효과적으로 처리할  있는 시스템개발이 가능해졌다현재 연구실에서는 DIS 위해 보다 정확한 손떨림으로 인한 움직임과 의도된 카메라의 움직임을 구분하는 연구  정확한 지역적전역적 움직임 백터를 예측하는 기법을 연구하고 있다.


Related Papers

[1] S.-J. Ko, S.-H. Lee, and K.-H. Lee, "Digital Image Stabilizing Algorithms based on Bit-plane Matching," IEEE Trans. Consum. Electron., vol. 44, no. 3, pp. 617-622, Aug. 1998.

[2] S.-J. Ko, S.-H. Lee, S.-W. Jeon, and E.-S. Kang, "Fast Digital Image Stabilizer Based on Gray-Coded Bit-Plane Matching," IEEE Trans. Consum. Electron., vol. 45, no. 3, pp. 598-603, Aug. 1999.

[3] S.-K. Kim, S.-J. Kang, T.-S. Wang, and S.-J. Ko, "Feature Point Classification based Global Motion Estimation for Video Stabilization," IEEE Trans. Consum. Electron., vol. 59, no. 1, pp. 267-272, Feb. 2013.


Related Patent

적응적인 칼만필터를 이용한 영상 안정화 장치  방법(Apparatus and method for image stabilization using adaptive Kalman filter), patent no. 10-0985805, Sep. 2010.

 

Flowchart of digital image stabilization (DIS)

image018.png


-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Current research topics:

- Fast Stabilization scheme to remove translational, rotational, and scaling disturbances [1]-[2]

- Motion estimation robust to the input image’s condition [3]

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------