• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.260-262
• /
• 2003

New Image registration techniques are developed for determining geometric distortions between two images of the same scene. First, the properties of the Fourier transform of a two dimensional function under the affine transformation are given. As a result, techniques for the estimation of the coefficients of the distortion model using the spectral frequency information are developed. Image registration can be achieved by applying the fast Fourier transform (FFT) technique for cross correlation of misregistered imagery to determine spatial distances. The correlation results may be rather broad, making detection of the peak difficult, what can be suppressed by enhancing cross-correlation technique. Yield greatly improves the delectability and high precision of image misregistration.

• Journal of Biomedical Engineering Research
• /
• v.30 no.1
• /
• pp.34-40
• /
• 2009

Transforms including translation and rotation are required for registering two or more images. In medical applications, different registration methods have been applied depending on the structures: for rigid bodies such as bone structures, affine transformation was widely used. In most previous research, a single transform was used for registering the whole images, which resulted in low registration accuracy especially when the degree of deformation was high between two images. In this paper, a novel registration method is introduced which is based image sub-division and bilinear interpolation of transformations. The proposed method enhanced the registration accuracy by 40% comparing with Trimmed ICP for registering color and MRI images.

• Journal of Korea Game Society
• /
• v.11 no.2
• /
• pp.161-166
• /
• 2011

In this paper, we propose a novel background segmentation and feature point extraction method of a human motion for the augmented reality game. First, our method transforms input image from RGB color space to HSV color space, then segments a skin colored area using double threshold of H, S value. And it also segments a moving area using the time difference images and then removes the noise of the area using the Hessian affine region detector. The skin colored area with the moving area is segmented as a human motion. Next, the feature points for the human motion are extracted by calculating the center point for each block in the previously obtained image. The experiments on various input images show that our method is capable of correct background segmentation and feature points extraction 12 frames per second.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.6C
• /
• pp.349-354
• /
• 2011

Passive millimeter wave imaging has the capability of detecting concealed objects under clothing. Also, passive millimeter imaging can obtain interpretable images under low visibility conditions like rain, fog, smoke, and dust. However, the image quality is often degraded due to low spatial resolution, low signal level, and low temperature resolution. This paper addresses image registration and fusion between passive millimeter images and visual images. The goal of this study is to combine and visualize two different types of information together: human subject's identity and concealed objects. The image registration process is composed of body boundary detection and an affine transform maximizing cross-correlation coefficients of two edge images. The image fusion process comprises three stages: discrete wavelet transform for image decomposition, a fusion rule for merging the coefficients, and the inverse transform for image synthesis. In the experiments, various types of metallic and non-metallic objects such as a knife, gel or liquid type beauty aids and a phone are detected by passive millimeter wave imaging. The registration and fusion process can visualize the meaningful information from two different types of sensors.

• Journal of Korea Multimedia Society
• /
• v.15 no.9
• /
• pp.1102-1111
• /
• 2012

In this paper, we propose a robust watermarking method against geometric attack even though the watermarked image is partially damaged. This method consists of standard image normalization which transforms any image into a predefined standard image and embedding watermark in DCT domain of standard normalized image using spread spectrum technique. The proposed standard image normalization method has an improvement over existing image normalization method, so it is robust to partial damage and geometric attack. The watermark embedding method using spread spectrum technique also has a robustness to image losses such as blurring, sharpening and compressions. In addition, the proposed watermarking method does not need an original image to detect watermark, so it is useful to public watermarking applications. Several experimental results show that the proposed watermarking method is robust to partial damage and various attacks including geometric deformation.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.26 no.2
• /
• pp.28-36
• /
• 2022

As port structures are exposed to various extreme external loads such as wind (typhoons), sea waves, or collision with ships; it is important to evaluate the structural safety periodically. To monitor the port structure, especially the rubber fender, a fender segmentation system using a vision sensor and deep learning method has been proposed in this study. For fender segmentation, a new deep learning network that improves the encoder-decoder framework with the receptive field block convolution module inspired by the eccentric function of the human visual system into the DenseNet format has been proposed. In order to train the network, various fender images such as BP, V, cell, cylindrical, and tire-types have been collected, and the images are augmented by applying four augmentation methods such as elastic distortion, horizontal flip, color jitter, and affine transforms. The proposed algorithm has been trained and verified with the collected various types of fender images, and the performance results showed that the system precisely segmented in real time with high IoU rate (84%) and F1 score (90%) in comparison with the conventional segmentation model, VGG16 with U-net. The trained network has been applied to the real images taken at one port in Republic of Korea, and found that the fenders are segmented with high accuracy even with a small dataset.

• 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.