• Smart Media Journal
• /
• v.9 no.4
• /
• pp.17-25
• /
• 2020

In the image object classification problem, low-resolution images may have a negative impact on the classification result, especially when the classification method, such as a convolutional neural network (CNN) model, is trained on a high-resolution (HR) image dataset. In this paper, we analyze the behavior of applying a classical super-resolution (SR) method such as bicubic interpolation, and a deep CNN model such as SRCNN to enhance low-resolution (LR) weeds images used for classification. Using an HR dataset, we first train a CNN model for weeds image classification with a default input size of 128 × 128. Then, given an LR weeds image, we rescale to default input size by applying the bicubic interpolation or the SRCNN model. We analyze these two approaches on the Chonnam National University (CNU) weeds dataset and find that SRCNN is suitable for the image size is smaller than 80 × 80, while bicubic interpolation is convenient for a larger image.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.8
• /
• pp.3942-3961
• /
• 2019

Recently, the models of deep super-resolution networks can successfully learn the non-linear mapping from the low-resolution inputs to high-resolution outputs. However, for large scaling factors, this approach has difficulties in learning the relation of low-resolution to high-resolution images, which lead to the poor restoration. In this paper, we propose Stage Generative Adversarial Networks (Stage-GAN) with semantic maps for image super-resolution (SR) in large scaling factors. We decompose the task of image super-resolution into a novel semantic map based reconstruction and refinement process. In the initial stage, the semantic maps based on the given low-resolution images can be generated by Stage-0 GAN. In the next stage, the generated semantic maps from Stage-0 and corresponding low-resolution images can be used to yield high-resolution images by Stage-1 GAN. In order to remove the reconstruction artifacts and blurs for high-resolution images, Stage-2 GAN based post-processing module is proposed in the last stage, which can reconstruct high-resolution images with photo-realistic details. Extensive experiments and comparisons with other SR methods demonstrate that our proposed method can restore photo-realistic images with visual improvements. For scale factor ${\times}8$, our method performs favorably against other methods in terms of gradients similarity.

• Journal of Internet Computing and Services
• /
• v.19 no.4
• /
• pp.45-52
• /
• 2018

A single-image super-resolution is a process of restoring a high-resolution image from a low-resolution image. Recently, the super-resolution using the deep neural network has shown good results. In this paper, we propose a neural network structure that improves speed and performance over conventional neural network based super-resolution methods. To do this, we analyze the conventional neural network based super-resolution methods and propose solutions. The proposed method reduce the 5 stages of the conventional method to 3 stages. Then we have studied the optimal width and depth by experimenting on the width and depth of the network. Experimental results have shown that the proposed method improves the disadvantages of the conventional methods. The proposed neural network structure showed superior performance and speed than the conventional method.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.8
• /
• pp.643-649
• /
• 2016

This paper presents a new image cache algorithm for real-time implementation of high-resolution color image warping. The cache memory is divided into four cache memory modules for simultaneous readout of four input image pixels in consideration of the color filter array (CFA) pattern of an image sensor and CFA image warping. In addition, a pipeline structure from the cache memory to an interpolator is shown to guarantee the generation of an output image pixel at each system clock cycle. The proposed image cache algorithm is applied to an FPGA-based real-time color image warping, and experimental results are presented to show the validity of the proposed method.

• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.865-869
• /
• 2009

An image resolution enhancement is mainly utilized as pre-processing technique for various image processing application. It requires to decrease image quality deterioration such as blurring. In this paper, we propose an image resolution enhancement algorithm using low level interpolation. In the proposed algorithm, we calculate an error using low level interpolation, estimate an error image from the calculated error. The estimated error image is added interpolated high resolution image, it become lastly reconstruction image. Our experiments obtained the average PSNR about 1dB which is improved results better than conventional method for sensitive image quality. Also, subjective image quality with edge region is more clearness. The proposed method may be helpful for applications in various multimedia systems such as image restoration.

• Journal of Broadcast Engineering
• /
• v.24 no.5
• /
• pp.703-712
• /
• 2019

Single image Super-Resolution (SISR) aims to generate a visually pleasing high-resolution image from its degraded low-resolution measurement. In recent years, deep learning - based super - resolution methods have been actively researched and have shown more reliable and high performance. A typical method is WaveletSRNet, which restores high-resolution images through wavelet coefficient learning based on feature maps of images. However, there are two disadvantages in WaveletSRNet. One is a big processing time due to the complexity of the algorithm. The other is not to utilize feature maps efficiently when extracting input image's features. To improve this problems, we propose an efficient single image super resolution method, named RDB-WaveletSRNet. The proposed method uses the residual dense block to effectively extract low-resolution feature maps to improve single image super-resolution performance. We also adjust appropriated growth rates to solve complex computational problems. In addition, wavelet packet decomposition is used to obtain the wavelet coefficients according to the possibility of large scale ratio. In the experimental result on various images, we have proven that the proposed method has faster processing time and better image quality than the conventional methods. Experimental results have shown that the proposed method has better image quality by increasing 0.1813dB of PSNR and 1.17 times faster than the conventional method.

• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 2006.06a
• /
• pp.34-39
• /
• 2006

An automated surface-strain measurement system, named ASIAS, was developed by using the image processing and stereo vision techniques in the previous studies by the corresponding author and his coworkers. This system has been upgraded mainly to improve the accuracy through image enhancement, sub-pixel measurement, surface smoothing, etc., since the first version was released. The present study has still more improved the convenience of users as well as the accuracy of measurement by processing high resolution images 8 mega pixels or more which can be easily obtained from a portable digital steal camera. It is proved that high resolution image processing greatly decreases the measurement error and gives strain data without considerable deterioration of accuracy even when the deformed grids to be measured and the master grids for camera calibration are captured together in the same image, making the whole process of strain measurement much simpler.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.1464-1466
• /
• 2003

classification of the satellite images provides information about land cover and/or land use. Quality of the classification result depends mainly on the spatial and spectral resolutions of the images. In this study, image fusion in terms of resolution merging, and band integration with multi-source of the satellite images; Landsat ETM+ and Ikonos were carried out to improve classification. Resolution merging and band integration could generate imagery of high resolution with more spectral bands. Precise image co-registration is required to remove geometric distortion between different sources of images. Combination of unsupervised and supervised classification of the fused imagery was implemented to improve classification. 3D display of the results was possible by combining DEM with the classification result so that interpretability could be improved.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.156-158
• /
• 2003

Analysis of high-resolution satellite image becomes important for cartography, surveillance, and remote sensing. However, there are lots of problems to be solved for automatic analysis of high-resolution satellite image especially in urban area. The problems are originated from the increased complexity due to the unnecessary details and shadows, and time-varying illuminations. Because of such obstacles, it seems impossible to make automatic analysis. This paper proposes a way of change detection of buildings in urban area by using digital vector map. The proposed way makes the buildings on the vector map parameterized, and searches them in the preprocessed high-resolution image by using generalized Hough transform. The searched building objects are overlaid on the satellite image. The overlaid image can help to detect the change of building rapidly.

• Korean Journal of Remote Sensing
• /
• v.18 no.6
• /
• pp.327-336
• /
• 2002

Ever since high resolution satellites were launched, high-resolution satellite images have been utilized in many areas. This paper proposed methods of generating simulated satellite image using DEM(Digital Elevation Model) and digital image such as aerial photograph. There are two methods proposed in the paper: one is Direct-Indirect method and the other Indirect-Indirect, method. It is assumed that satellite attitude is not changing and perspective center is moving in the direction of flight while image is captured. The proposed methods were implemented with aerial photograph, DEM data, arbitrary orbit parameters and attitude parameters of high resolution satellite image under generation. Furthermore, for the stereo viewing, different orientation parameters and perspective center were tested for generating simulated satellite image. In addition, the quality and accuracy of the simulated satellite image generated by the proposed methods were analyzed.