• 한국멀티미디어학회논문지
• /
• 제21권12호
• /
• pp.1467-1472
• /
• 2018

For the problem that the face image of surveillance video cannot be accurately identified due to the low resolution, this paper proposes a low resolution face recognition solution based on convolutional neural network model. Convolutional Neural Networks (CNN) model for multi-scale input The CNN model for multi-scale input is an improvement over the existing "two-step method" in which low-resolution images are up-sampled using a simple bi-cubic interpolation method. Then, the up sampled image and the high-resolution image are mixed as a model training sample. The CNN model learns the common feature space of the high- and low-resolution images, and then measures the feature similarity through the cosine distance. Finally, the recognition result is given. The experiments on the CMU PIE and Extended Yale B datasets show that the accuracy of the model is better than other comparison methods. Compared with the CMDA_BGE algorithm with the highest recognition rate, the accuracy rate is 2.5%~9.9%.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.517-519
• /
• 2004

This paper describes an approach to estimate a robot pose with an image. The algorithm of pose estimation with an image can be broken down into three stages : extracting scale-invariant features, matching these features and calculating affine invariant. In the first step, the robot mounted mono camera captures environment image. Then feature extraction is executed in a captured image. These extracted features are recorded in a database. In the matching stage, a Random Sample Consensus(RANSAC) method is employed to match these features. After matching these features, the robot pose is estimated with positions of features by calculating affine invariant. This algorithm is implemented and demonstrated by Matlab program.

• 대한의용생체공학회:학술대회논문집
• /
• 대한의용생체공학회 1995년도 추계학술대회
• /
• pp.113-116
• /
• 1995

Medical image compression using the wavelet transform has been tried. Due to the flexibility in representing nonstationary image signal in both time and frequency domains and its ability to adapt human visual characteristics, wavelet transform has unique advantage in images compression. In the proposed wavelet compression original image is decomposed into multi-scale bands. Different scale factors are employed in the quantization of wavelet decomposed images in different bands. For the lowest band, a predictor is designed and error signal is entropy coded. For high scale bands, runlength coding for toro run is used with Huffman coding. From simulation with magnetic resonance images($256\times256$ size, 256 graylevels) the proposed algorithm is superior to the JPEG by more than 2.5 dB in near lossless compression (CR = 8 - 10).

• Journal of information and communication convergence engineering
• /
• 제10권2호
• /
• pp.194-199
• /
• 2012

This paper proposes a novel scheme of a gray scale fingerprint image for a high-accuracy capacitive sensor chip. The conventional grayscale image scheme uses a digital-to-analog converter (DAC) of a large-scale layout or charge-pump circuit with high power consumption and complexity by a global clock signal. A modified capacitive detection circuit for the charge sharing scheme is proposed, which uses a down literal circuit (DLC) with a floating-gate metal-oxide semiconductor transistor (FGMOS) based on a neuron model. The detection circuit is designed and simulated in a 3.3 V, 0.35 ${\mu}m$ standard CMOS process. Because the proposed circuit does not need a comparator and peripheral circuits, the pixel layout size can be reduced and the image resolution can be improved.

• 천문학회지
• /
• 제52권2호
• /
• pp.49-55
• /
• 2019

The New Vacuum Solar Telescope (NVST) is the largest solar telescope in China. When using CCDs for imaging, equal-thickness fringes caused by thin-film interference can occur. Such fringes reduce the quality of NVST data but cannot be removed using standard flat fielding. In this paper, a correction method based on multi-scale decomposition and adaptive partitioning is proposed. The original image is decomposed into several sub-scales by multi-scale decomposition. The region containing fringes is found and divided by an adaptive partitioning method. The interference fringes are then filtered by a frequency-domain Gaussian filter on every partitioned image. Our analysis shows that this method can effectively remove the interference fringes from a solar image while preserving useful information.

• 스마트미디어저널
• /
• 제12권9호
• /
• pp.45-59
• /
• 2023

Weakly supervised object localization (WSOL) is a task of localizing an object in an image using only image-level labels. Previous studies have followed the conventional class activation mapping (CAM) pipeline. However, we reveal the current CAM approach suffers from problems which cause original CAM could not capture the complete defects features. This work utilizes a convolutional neural network (CNN) pretrained on image-level labels to generate class activation maps in a multi-scale manner to highlight discriminative regions. Additionally, a vision transformer (ViT) pretrained was treated to produce multi-head attention maps as an auxiliary detector. By integrating the CNN-based CAMs and attention maps, our approach localizes defective regions without requiring bounding box or pixel-level supervision during training. We evaluate our approach on a dataset of apple images with only image-level labels of defect categories. Experiments demonstrate our proposed method aligns with several Object Detection models performance, hold a promise for improving localization.

• 한국산학기술학회논문지
• /
• 제15권9호
• /
• pp.5750-5757
• /
• 2014

본 논문은 의도하지 않게 카메라에 입력되는 회전, 병진 움직임뿐만 아니라 확대 축소에 의한 움직임을 SIFT 기법을 이용하여 추정하고 제거하는 영상 안정화 기법을 제시한다. 또한 영상 분할 및 통합 방법을 이용하여 움직이는 물체가 영상에 입력되었을 경우에도 이러한 외란을 제거하고 영상 안정화를 수행하도록 한다. 제시된 방법은 다른 방법과 결과를 비교하는 실험을 수행하여 성능이 우수함을 검증한다.

• Journal of information and communication convergence engineering
• /
• 제10권1호
• /
• pp.45-52
• /
• 2012

Thedenoising of a natural image corrupted by Gaussian noise is a long established problem in signal or image processing. Today the research is focus on the wavelet domain, especially using the wavelet threshold method. In this paper, a waveletbased image denoising modified adaptive thresholding method is proposed. The proposed method computes thethreshold adaptively based on the scale level and adaptively estimates wavelet coefficients by using a modified thresholding function that considers the dependency between the parent coefficient and child coefficient and the soft thresholding function at different scales. Experimental results show that the proposed method provides high peak signal-to-noise ratio results and preserves the detailed information of the original image well, resulting in a superior quality image.

• 대한전자공학회논문지
• /
• 제26권10호
• /
• pp.1554-1560
• /
• 1989

This paper suggests the algorihtm which extracts the classification parameter relative to the only feature of document patterns even though they are rotated or scaled, and also classifies them. With the complex logarithmic conformal mapping, the sample of the document pattern image makes the pattern image of the complex logarithmic plane. Because the power spectrum of this plane is invariant to the rotation, and scale of the pattern image, it is used as the characteristics parameter of the patten image. By using the coherence function, this method analyzes the standard and input power spectrum. additionally, it classifies the input pattern image. Even though input image is rotated, our algorithm can classify it without reference to the rotation, and this is possible when the scale is in the range of 0.5-1.5.

• 한국멀티미디어학회논문지
• /
• 제20권9호
• /
• pp.1481-1490
• /
• 2017

This paper presents an image quality enhancement algorithm for dark image acquired under poor lighting condition. Various retinex algorithms which are human perception-based image processing methods were proposed to solve this problem. Although MSR(Multi-Scale Retinex) among these algorithm works well under most lighting condition, it shows color degradation because their separate nonlinear processing of RGB color channels. To compensate for the loss of the color, MSRCR(Multi-Scale Retinex with Color Restoration) was proposed. However, it requires high computational load and has additional parameters that need to be adjusted according to input image. In order to overcome this problem, a new retinex algorithm based on MSR is proposed in this paper. The proposed method consists of V channel MSR, saturation correction, and separate contrast enhancement process. Experimental results show that the subjective and objective image quality of the proposed method better than those of the conventional methods.