Mandibular fracture is one of the most frequent injuries in oral and maxillo-facial surgery. Radiologists diagnose mandibular fractures using panoramic radiography and cone-beam computed tomography (CBCT). Panoramic radiography is a conventional imaging modality, which is less complicated than CBCT. This paper proposes the diagnosis method of mandibular fractures in a panoramic radiograph based on a deep learning system without the intervention of radiologists. The deep learning system used has a one-stage detection called you only look once (YOLO). To improve detection accuracy, panoramic radiographs as input images are augmented using gamma modulation, multi-bounding boxes, single-scale luminance adaptation transform, and multi-scale luminance adaptation transform methods. Our results showed better detection performance than the conventional method using YOLO-based deep learning. Hence, it will be helpful for radiologists to double-check the diagnosis of mandibular fractures.

High dynamic range (HDR) imaging refers to digital image processing that modifies the range of color and contrast to enhance image visibility. To create an HDR image, two or more images that include various information are needed. In order to convert low dynamic range (LDR) images to HDR images, we consider the possibility of using a generative adversarial network (GAN) as an appropriate deep neural network. Deep learning requires a great deal of data in order to build a module, but once the module is created, it is convenient to use. In this paper, we propose a weight map for local luminance based on learning to reconstruct locally tone-mapped images.

Mandibular fractures are the most common fractures in dentistry. Since diagnosing a mandibular fracture is difficult when only panoramic radiographic images are used, most doctors use cone beam computed tomography (CBCT) to identify the patient’s fracture location. In this study, considering the diagnosis of mandibular fractures using the combined deep learning technique, YOLO and U-Net were used as auxiliary diagnostic methods to detect the location of mandibular fractures based on panoramic images without CBCT. In a previous study, mandibular fracture diagnosis was performed using YOLO learning; in the detection performance result of the YOLOv4-based mandibular fracture diagnosis module, the precision score was approximately 97%, indicating that there was almost no misdiagnosis. In particular, fractures in the symphysis, body, angle, and ramus tend to be distributed in the middle of the mandible. Owing to the irregular fracture types and overlapping location information, the recall score was approximately 79%, which increased the detection of undiagnosed fractures. In many cases, fractures that are clearly visible to the human eye cannot be grasped. To overcome these shortcomings, the number of undiagnosed fractures can be reduced using a combination of the U-Net and YOLOv4 learning modules. U-Net is advantageous for the segmentation of fractures spread over a wide area because it performs semantic segmentation. Consequently, the undiagnosed case in the middle of the mandible, where YOLO was weak, was somewhat supplemented by the U-Net module. The precision score of the combined module was 95%, similar to that of the previous method, and the recall score improved to 87%, as the number of undiagnosed cases was reduced. Through this study, the performance of a deep learning method that can be used for the diagnosis of the mandibular bone has been improved, and it is anticipated that as an auxiliary diagnostic inspection device, it will assist dentists in making diagnoses.

This study proposes a method of blending visible and near-infrared (NIR) images to enhance their edge details and local contrast based on the Laplacian pyramid and principal component analysis (PCA). In the proposed method, both the Laplacian pyramid and PCA are implemented to generate a radiance map. Using the PCA algorithm, the soft-mixing method and the mask-skipping filter were applied when the images were fused. The color compensation method uses the ratio between the radiance map fused by the Laplacian pyramid and the PCA algorithm and the luminance channel of the visible image to preserve the chrominance of the visible image. The results show that the proposed method improves edge details and local contrast effectively.

No abstract

This study aims to develop a spatial dual-sensor module for acquiring visible and near-infrared images in the same space without time shifting and to synthesize the captured images. The proposed method synthesizes visible and near-infrared images using contourlet transform, principal component analysis, and iCAM06, while the blending method uses color information in a visible image and detailed information in an infrared image. The contourlet transform obtains detailed information and can decompose an image into directional images, making it better in obtaining detailed information than decomposition algorithms. The global tone information is enhanced by iCAM06, which is used for high-dynamic range imaging. The result of the blended images shows a clear appearance through both the compressed tone information of the visible image and the details of the infrared image.

In this paper, to improve the slow processing speed of the rule-based visible and NIR (near-infrared) image synthesis method, we present a fast image fusion method using DenseFuse, one of the CNN (convolutional neural network)-based image synthesis methods. The proposed method applies a raster scan algorithm to secure visible and NIR datasets for effective learning and presents a dataset classification method using luminance and variance. Additionally, in this paper, a method for synthesizing a feature map in a fusion layer is presented and compared with the method for synthesizing a feature map in other fusion layers. The proposed method learns the superior image quality of the rule-based image synthesis method and shows a clear synthesized image with better visibility than other existing learning-based image synthesis methods. Compared with the rule-based image synthesis method used as the target image, the proposed method has an advantage in processing speed by reducing the processing time to three times or more.

High-dynamic-range (HDR) image synthesis is a technology developed to accurately reproduce the actual scene of an image on a display by extending the dynamic range of an image. Multi-exposure fusion (MEF) technology, which synthesizes multiple low-dynamic-range (LDR) images to create an HDR image, has been developed in various ways including pixel-based, patch-based, and deep learning-based methods. Recently, methods to improve the synthesis quality of images using deep-learning-based algorithms have mainly been studied in the field of MEF. Despite the various advantages of deep learning, deep-learning-based methods have a problem in that numerous multi-exposed and ground-truth images are required for training. In this study, we propose a self-supervised learning method that generates and learns reference images based on input images during the training process. In addition, we propose a method to train a deep learning model for an MEF with multiple tasks using dynamic hyperparameters on the loss functions. It enables effective network optimization across multiple tasks and high-quality image synthesis while preserving a simple network architecture. Our learning method applied to the deep learning model shows superior synthesis results compared to other existing deep-learning-based image synthesis algorithms.