• Journal of Conservation Science
• /
• v.38 no.1
• /
• pp.14-32
• /
• 2022

Underwater photographing and image recording are essential for pre-excavation survey and during excavation in underwater archaeology. Unlike photographing on land, all underwater images suffer various quality degradations such as shape distortions, color shift, blur, low contrast, high noise levels and so on. Outcome is very often heavily photographing equipment and photographer dependent. Excavation schedule, weather conditions, and water conditions can put burdens on divers. Usable images are very limited compared to the efforts. In underwater archaeological study in very turbid water such as in the Yellow Sea (between mainland China and the Korean peninsula), underwater photographing is very challenging. In this study, off-site image distortion and color compensation techniques using an image processing/analysis software is investigated as an alternative image quality enhancement method. As sample images, photographs taken during the excavation of 800-year-old Taean Mado Shipwrecks in the Yellow Sea in 2008-2010 were mainly used. Significant enhancement in distortion and color compensation of archived images were obtained by simple post image processing using image processing/analysis software (PicMan) customized for given view ports, lenses and cameras with and without optical axis offsets. Post image processing is found to be very effective in distortion and color compensation of both recent and archived images from various photographing equipment models and configurations. Merits and demerit of in-situ, distortion and color compensated photographing with sophisticated equipment and conventional photographing equipment, which requires post image processing, are compared.

• Journal of Broadcast Engineering
• /
• v.28 no.2
• /
• pp.178-184
• /
• 2023

Finding texts in general scene images and recognizing their contents is a very important task that can be used as a basis for robot vision, visual assistance, and so on. However, for the low-resolution text images, the degradations, such as noise or blur included in text images, are more noticeable, which leads to severe performance degradation of text recognition accuracy. In this paper, we propose a new Korean text image super-resolution based on a Transformer-based model, which generally shows higher performance than convolutional neural networks. In the experiments, we show that text recognition accuracy for Korean text images can be improved when our proposed text image super-resolution method is used. We also propose a new Korean text image dataset for training our model, which contains massive HR-LR Korean text image pairs.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.9
• /
• pp.127-135
• /
• 2013

In order to acquire images in low-light environments, it is usually necessary to adopt long exposure times or resort to flash lights. However, flashes often induce color distortion, cause the red-eye effect and can be disturbing to subjects. On the other hand, long-exposure shots are susceptible to subject-motion, as well as motion-blur due to camera shake when performed hand-held. A recently introduced technique to overcome the limitations of traditional low-light photography is that of multi-spectral flash. Multi-spectral flash images are a combination of UV/IR and visible spectrum information. The general idea is that of retrieving details from the UV/IR spectrum and color from the visible spectrum. However, multi-spectral flash images themselves are subject to color distortion and noise. This works presents a method to compute multi-spectral flash images so that noise can be reduced and color accuracy improved. The proposed approach is a previously seen optimization method, improved by the introduction of a weight map used to discriminate uniform regions from detail regions. The weight map is generated by applying canny edge operator and it is applied to the optimization process for discriminating the weights in uniform region and edge. Accordingly, the weight of color information is increased in the uniform region and the detail region of weight is decreased in detail region. Therefore, the proposed method can be enhancing color reproduction and removing artifacts. The performance of the proposed method has been objectively evaluated using long-exposure shots as reference.

• Progress in Medical Physics
• /
• v.23 no.1
• /
• pp.26-32
• /
• 2012

The magnification technique has recently become popular in bone radiography, mammography and other diagnostic examination. However, because of the finite size of X-ray focal spot, the magnification influences various imaging properties with resolution, noise and contrast. The purpose of study is to investigate the influence of magnification and focal spot size on digital imaging system using eDQE (effective detective quantum efficiency). Effective DQE is a metric reflecting overall system response including focal spot blur, magnification, scatter and grid response. The adult chest phantom employed in the Food and Drug Administration (FDA) was used to derive eDQE from eMTF (effective modulation transfer function), eNPS (effective noise power spectrum), scatter fraction and transmission fraction. According to results, spatial frequencies that eMTF is 10% with the magnification factor of 1.2, 1.4, 1.6, 1.8 and 2.0 are 2.76, 2.21, 1.78, 1.49 and 1.26 lp/mm respectively using small focal spot. The spatial frequencies that eMTF is 10% with the magnification factor of 1.2, 1.4, 1.6, 1.8 and 2.0 are 2.21, 1.66, 1.25, 0.93 and 0.73 lp/mm respectively using large focal spot. The eMTFs and eDQEs decreases with increasing magnification factor. Although there are no significant differences with focal spot size on eDQE (0), the eDQEs drops more sharply with large focal spot than small focal spot. The magnification imaging can enlarge the small size lesion and improve the contrast due to decrease of effective noise and scatter with air-gap effect. The enlargement of the image size can be helpful for visual detection of small image. However, focal spot blurring caused by finite size of focal spot shows more significant impact on spatial resolution than the improvement of other metrics resulted by magnification effect. Based on these results, appropriate magnification factor and focal spot size should be established to perform magnification imaging with digital radiography system.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.5
• /
• pp.84-95
• /
• 2011

In this paper, we present a deblocking algorithm that removes grid and staircase noises, which are called "blocking artifacts", occurred in the compressed satellite images. Particularly, the given satellite images are compressed with equal quantization coefficients in row according to region complexity, and more complicated regions are compressed more. However, this approach has a problem that relatively less complicated regions within the same row of complicated regions have blocking artifacts. Removing these artifacts with a general deblocking algorithm can blur complex and undesired regions as well. Additionally, the general filter lacks in preserving the curved edges. Therefore, the proposed algorithm presents an adaptive filtering scheme for removing blocking artifacts while preserving the image details including curved edges using the given quantization step size and content analysis. Particularly, WLFPCA (weighted lowpass filter using principle component analysis) is employed to reduce the artifacts around edges. Experimental results showed that the proposed method outperforms SA-DCT in terms of subjective image quality.

• KIPS Transactions on Software and Data Engineering
• /
• v.1 no.2
• /
• pp.103-108
• /
• 2012

Recently, as the prevalence of high-performance mobile devices and the application of the multimedia content are expanded, Super Resolution (SR) technique which reconstructs low resolution images to high resolution images is becoming important. And in the mobile devices, the development of the SR algorithm considering the operation quantity or memory is required because of using the restricted resources. In this paper, we propose a new single frame fast SR technique suitable for mobile devices. In order to prevent color distortion, we change RGB color domain to HSV color domain and process the brightness information V (Value) considering the characteristics of human visual perception. First, the low resolution image is enlarged by the improved fast back projection considering the noise elimination. And at the same time, the reliable edge map is extracted by using the LoG (Laplacian of Gaussian) filtering. Finally, the high definition picture is reconstructed by using the edge information and the improved back projection result. The proposed technique removes effectually the unnatural artefact which is generated during the super resolution restoration, and the edge information which can be lost is amended and emphasized. The experimental results indicate that the proposed algorithm provides better performance than conventional back projection and interpolation methods.

• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.453-462
• /
• 2021

A lot of license plate data is required for car number recognition. License plate data needs to be balanced from past license plates to the latest license plates. However, it is difficult to obtain data from the actual past license plate to the latest ones. In order to solve this problem, a license plate recognition study through deep learning is being conducted by creating a synthetic license plates. Since the synthetic data have differences from real data, and various data augmentation techniques are used to solve these problems. Existing data augmentation simply used methods such as brightness, rotation, affine transformation, blur, and noise. In this paper, we apply a style transformation method that transforms synthetic data into real-world data styles with data augmentation methods. In addition, real license plate data are noisy when it is captured from a distance and under the dark environment. If we simply recognize characters with input data, chances of misrecognition are high. To improve character recognition, in this paper, we applied the DeblurGANv2 method as a quality improvement method for character recognition, increasing the accuracy of license plate recognition. The method of deep learning for license plate detection and license plate number recognition used YOLO-V5. To determine the performance of the synthetic license plate data, we construct a test set by collecting our own secured license plates. License plate detection without style conversion recorded 0.614 mAP. As a result of applying the style transformation, we confirm that the license plate detection performance was improved by recording 0.679mAP. In addition, the successul detection rate without image enhancement was 0.872, and the detection rate was 0.915 after image enhancement, confirming that the performance improved.