의사 참조 이미지를 활용한 무참조 이미지 화질 평가

용윤정,오희석 대한전자공학회 2024 전자공학회논문지 Vol.61 No.2

No-reference image quality assessment (NR-IQA) aims to objectively quantify the level of image quality degradation by reflecting the human visual system in the absence of information about the pristine image. Existing NR-IQA techniques have high sensitivity to specific distortion types, but have limitations in determining semantic quality information or global image quality degradation. In this paper, to resolve the limitations of existing NR-IQA approaches and improve predictive power, we propose a multi-scale pseudo image quality assessor (MPIQ). MPIQ is an NR-IQA model that follows the framework of full-reference IQA (FR-IQA), which is more proficient in extracting local distortion patterns and aggregating higher-level perceptual quality information. The proposed MPIQ employs a hybrid scheme that seeks to understand local distortion patterns through the convolutional neural networks and global level of image quality based on transformers, and consists of two modules: a pseudo-reference image reconstructor and a quasi FR-IQA regressor. Similar to the FR-IQA approach, the pseudo-reference image reconstructor utilizes an encoder-decoder structure to reconstruct the pseudo-reference image and learn image degradation information compared to a distorted one. Here, a multi-scale structure is reflected to extract scale-invariant distortion patterns. Quasi FR-IQA regressor predicts the image quality score by deriving the global distortion level through the difference between the features extracted from the distorted and the pseudo-reference images. MPIQ was supervised onto the mean opinion score obtained through subjective evaluation in an end-to-end manner, and experimental results showed a 20% performance improvement compared to the existing NR-IQA. 무참조 이미지 화질 평가(no-reference image quality assessment; NR-IQA)는 참조 영상에 대한 정보가 주어지지 않은 상태에서 인간 시각 체계를 반영하여 이미지 화질 저하 수준을 객관적으로 정량화 하는 것을 목표로 한다. 기존 NR-IQA 기술은 특정 왜곡 특성에 대해 높은 민감성을 갖고 있으며, 의미론적 화질 인지 및 화질 저하 수준을 판단하기에는 한계를 지니고 있다. 본 논문에서는 기존 NR-IQA 접근 방식의 한계를 극복하고 예측 정확도를 향상시키기 위해 지역적 왜곡 패턴 추출과 고수준 의미 정보 추출에 보다 능숙한 완전 참조 (full-reference IQA; FR-IQA) 방식의 프레임워크를 따르는 NR-IQA 모델인 multi-scale pseudo image quality assessor(MPIQ)를 제안한다. 제안하는 MPIQ는 하이브리드 구조에 기반해 합성곱 네트워크를 통한 지역적 왜곡 패턴 추출과 트랜스포머 기반의 전역적 화질 수준 이해를 도모하며, 의사 참조 이미지 reconstructor와 quasi FR-IQA regressor라는 두 가지 모듈로 구성된다. 의사 참조 이미지 reconstructor는 FR-IQA 방식과 유사하게 화질 추론 모델이 동작할 수 있도록 인코더-디코더 구조를 차용하여 의사 참조 이미지를 재구성하고 이미지 왜곡 정보를 학습한다. 이 과정에서 크기 불변 왜곡 패턴의 추출을 위해 다중 스케일 구조를 반영하였다. Quasi FR-IQA regressor는 왜곡 이미지에서 추출된 특징과 의사 참조 이미지 특징의 차이를 통해 전역적 왜곡 수준을 도출함으로써 이미지 화질 수준을 최종적으로 예측한다. MPIQ는 주관적 평가를 통해 획득된 mean opinion score에 end-to-end로 학습되었으며, 실험 결과 기존 NR-IQA 기술 대비 20%의 성능 향상을 달성하였다.

Research on Camera Calibration method Based on Eight Parameter Model

Yu Shuchun,Liu Shuang,Yu Xiaoyang 보안공학연구지원센터 2015 International Journal of Signal Processing, Image Vol.8 No.9

Camera calibration is the basis for realization of image distortion correction. Because the causes of image distortion are very complex, it has become a difficulty to correct image distortion with calibration results. In order to reduce the complexity of the existing distortion correction method, an eight parameters camera calibration model was proposed in this paper. In this model, eight parameters were used in describing radial distortion and tangential distortion. The experimental results show that the proposed method can be easily solved and correct distortion accurately.

HUD를 위한 영상 보간과 보색 알고리즘 구현

안명수(Ming-Shou An),강대성(Dae-Seong Kang) 한국정보기술학회 2009 한국정보기술학회논문지 Vol.7 No.2

In this paper, we study distortion correction and complementary color extraction algorithm for the image of HUD(Head-Up Display). Complementary color extraction algorithm extracts the color of symmetrical location in complementary color coordinates using RGB value based on HSI color model. The main reason of image distortion in HUD that is not orthogonal between the front glass of car and the projected image. In this case, the projected image on the front glass of car is turned into the distortion image through projective transformation. Therefore to be restored the original image, the distortion image is projected on the front glass of car with a contrast to the distortion rate between the original and distortion image. And then, we get the contrary rate of image from distortion image with these parameters. In experiments, we propose the method using B-spline interpolation, and compare with previous method with projection transformation to show superiority.

Study on Distortion Compensation of Underwater Archaeological Images Acquired through a Fisheye Lens and Practical Suggestions for Underwater Photography - A Case of Taean Mado Shipwreck No. 1 and No. 2 -

Jung, Young-Hwa,Kim, Gyuho,Yoo, Woo Sik The Korean Society Of Conservation Science For Cul 2021 보존과학회지 Vol.37 No.4

Underwater archaeology relies heavily on photography and video image recording during surveillances and excavations like ordinary archaeological studies on land. All underwater images suffer poor image quality and distortions due to poor visibility, low contrast and blur, caused by differences in refractive indices of water and air, properties of selected lenses and shapes of viewports. In the Yellow Sea (between mainland China and the Korean peninsula), the visibility underwater is far less than 1 m, typically in the range of 30 cm to 50 cm, on even a clear day, due to very high turbidity. For photographing 1 m x 1 m grids underwater, a very wide view angle (180°) fisheye lens with an 8 mm focal length is intentionally used despite unwanted severe barrel-shaped image distortion, even with a dome port camera housing. It is very difficult to map wide underwater archaeological excavation sites by combining severely distorted images. Development of practical compensation methods for distorted underwater images acquired through the fisheye lens is strongly desired. In this study, the source of image distortion in underwater photography is investigated. We have identified the source of image distortion as the mismatching, in optical axis and focal points, between dome port housing and fisheye lens. A practical image distortion compensation method, using customized image processing software, was explored and verified using archived underwater excavation images for effectiveness in underwater archaeological applications. To minimize unusable area due to severe distortion after distortion compensation, practical underwater photography guidelines are suggested.

딥러닝 기반의 2차원 이미지 왜곡 분류 및 이미지 각도 보정 시스템 개발

서창진,김선화,김유진,최수민 대한전기학회 2022 전기학회논문지 P Vol.71 No.3

When taking a picture with a camera, the distortion that is different from reality occurs due to wide-angle lenses and camera angles. In this paper, we propose an image distortion classification and calibration program that provides users with standard images before distortion by classification and calibrating distortion. The program automatically predicts camera parameters from a single input image and proceeds with calibration. Inputting the image, distortion image classification using deep learning (CNN) determines whether Wide-angle lens distortion and Camera-angle distortion exist. When it is determined that distortion exists, deep learning and OpenCV are used to calibrate the distortion state according to each image characteristic. As a result of the program operation, it was confirmed that the output image was calibrated similarly to the actual one, and more fine distortion calibration can be expected by finding distortions that were difficult to judge only with human eyes.

정형보철용 금속이 자기공명영상에 미치는 영향Ⅰ

김형균(Hyeong-Gyun Kim),최성대(Seong-Dae Choi) 한국기계가공학회 2012 한국기계가공학회지 Vol.11 No.4

Stainless steel and titanium used for orthopedic prosthetic metal, magnetic resonance imaging was measured quantitatively the extent of distortion. Based on reading tests at varying metal magnetic resonance imaging (MRI) were investigated. To two image acquisition conditions for having three metal bodies mutually quantitative it compared analyzed the image of cross direction and the image of lengthwise. As for the distortion of the magnetic resonance image, as for the distortion of image it became small in the order of Clip, Strainless and Titanium. In addition, with T1WI and T2WI which are image acquisition condition T2WI the distortion of image became small. As for the distortion of the image by the metal, you could see the distortion of image in elliptical shape, the metal a bigger image distortion appeared in many parts. If clinical doctor has the condition of the same operation, use the Titanium material, it can, raising the diagnostic value of magnetic resonance image inspection after the operating from Stainless.

Research on Unwrapping Algorithm of Concentric Annular Image

Zhang Sheng,Lin Ming-xing,Wang Qing-dong 보안공학연구지원센터 2016 International Journal of Signal Processing, Image Vol.9 No.11

To meet the demand of rapid, accurate and online inspection of precision parts’ defects, the CCD camera is used to collect information of the inner hole. Since the obtained source image is a concentric annular image of 360 degrees with large image distortion, meanwhile the collected source images are not suitable to be processed directly with the image processing technology. So further processing is needed to be done for the rectangle unwrapping images. The method of Hough transform is adopted in this paper to find the center of a circle and the corresponding position used the center as a reference. On the basis of positioning the circle center, a kind of circular image unwrapping algorithm is designed by the transformational relation between Cartesian and polar coordinate. During unwrapping of the image, the bilinear interpolation algorithm is adopted to improve the image quality when the image unwrapping algorithm is used. The algorithm is verified by program and the result indicates that the algorithm to unwrap the image is efficient in locating the center by Hough transform combined with the coordinate’s transformation and the bilinear interpolation. Furthermore, the algorithm can unfold the image accurately and quickly and ensure the quality of the image, which prepares the foundation for subsequent processing.

자기공명영상을 이용한 근육 단면적 측정법의 활용을 위한 영상왜곡보정

홍철표,이동훈,박지원,한봉수,Hong, Cheol-Pyo,Lee, Dong-Hoon,Park, Ji-Won,Han, Bong-Soo 대한물리치료학회 2012 대한물리치료학회지 Vol.24 No.2

Purpose: The purpose of this study is to explore the importance of the image distortion correction in the cross sectional area measurement for the iliopsas muscle, tensor fasciae latae muscle, gluteus maximus muscle and the knee extensor muscles, by using (magnetic resonance imaging) MRI. Methods: This study was performed using an open 0.32T MRI system. To estimate the image distortion, T1 images for an AAPM homogeneity/linearity phantom were acquired, and the region in which the maximum geometric distortion was less than or equal to the pixel size (1.6 mm) of the images, it was defined as the distortion correction-free region. The T2 images for a human subject's pelvis and thigh in normal positions were obtained. Then, after the regions of interest in the pelvis and thigh were moved into the distortion correction-free region, T2 images for the pelvis and thigh were scanned with the same imaging parameters used in the previous T2 imaging. The cross-sectional areas were measured in the two T2 images that were obtained in the normal position, and the distortion correction-free region, as well as the area error caused by geometric image distortion was calculated. Results: The geometrical distortion is gradually increased, from the magnet center to the outer region, in axial and coronal plane. The cross-sectional area error of gluteus maximus muscle and the knee extensors was as high as 9.27% and 3.16% in before and after distortion correction, respectively. Conclusion: The cross-sectional area of the muscles that suffered from the geometrical distortion is necessary to correct for the estimation of the intervention.

연구논문 : 여대생들의 신체이미지 인식과 불만족 관련요인

김희진 ( Hee Jin Kim ),김창국 ( Chang Kuk Kim ) 한국미용학회 2006 한국미용학회지 Vol.12 No.3

The purpose of this study was to examine and investigate, the self body image perception and dissatisfaction of female college students and related factors. The data were collected and analyzed from 142 female college students using a structured and self-administrated questionnaire. The following results were obtained. The percentages of female college students distorted body image perception were 57.6%, and normal body image perception were 42.4%. As Contour Drawing Rating Scale, the mean level of current body image scale were 5.28, the mean level of ideal body image scale were 3.70 and mean body image dissatisfaction were 1.68, each. Significantly different factors between normal and distorted body image perception group were body image dissatisfaction and self out figure interest. As multiple linear regression results, the factors significantly related to body image dissatisfaction of distorted body image perception group included higher level of self-obesity perception, depression, and self out figure interest. As performing a health education about body image and obesity, there are some essential points for correcting distorted body image. Obesity woman is valuated not by appearance, but by ability, and not by outer-figure, but by inner-figure. Woman of distorted body image perception should be educated to perceive their real body image by BMI and other scientific indices. Surrounding persons should be educated as well as a subject of education.

안개가 포함된 영상에서의 색 왜곡 특성 분석

김정엽 아이씨티플랫폼학회 2023 JOURNAL OF PLATFORM TECHNOLOGY Vol.11 No.6

In this paper, the color distortion in images with haze would be analyzed. When haze is included in the scene, the color signal reflected in the scene is accompanied by color distortion due to the influence of transmittance according to the haze component. When the influence of haze is excluded by a conventional de-hazing method, the distortion of color tends to not be sufficiently resolved. Khoury et al. used the dark channel priority technique, a haze model mentioned in many studies, to determine the degree of color distortion. However, only the tendency of distortion such as color error values was confirmed, and specific color distortion analysis was not performed. This paper analyzes the characteristic of color distortion and proposes a restoration method that can reduce color distortion. Input images of databases used by Khoury et al. include Macbeth color checker, a standard color tool. Using Macbeth color checker's color values, color distortion according to changes in haze concentration was analyzed, and a new color distortion model was proposed through modeling. The proposed method is to obtain a mapping function using the change in chromaticity by step according to the change in haze concentration and the color of the ground truth. Since the form of color distortion varies from step to step in proportion to the haze concentration, it is necessary to obtain an integrated thought function that operates stably at all stages. In this paper, the improvement of color distortion through the proposed method was estimated based on the value of angular error, and it was verified that there was an improvement effect of about 15% compared to the conventional method.