• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.2 no.3
• /
• pp.39-52
• /
• 2009

Scale Invariant Feature Transform (SIFT) is an effective algorithm in object recognition, panorama stitching, and image matching, however, due to its complexity, real time processing is difficult to achieve with software approaches. This paper proposes using a reconfigurable hardware processor with integer half kernel. The integer half kernel Gaussian reduces the Gaussian pyramid complexity in about half [] and the reconfigurable processor carries out a parallel implementation of a full search Fast SIFT algorithm. We use a low memory, fine grain single instruction stream multiple data stream (SIMD) pixel processor that is currently being developed. This implementation fully exposes the available parallelism of the SIFT algorithm process and exploits the processing and I/O capabilities of the processor which results in a system that can perform real time image and video compression. We apply this novel implementation to images and measure the effectiveness. Experimental simulation results indicate that the proposed implementation is capable of real time applications.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 1999.10a
• /
• pp.321-332
• /
• 1999

It is important to detect objects movement for obstacle recognition and path searching of UCT(unmanned container transporters) with vision sensor. This paper shows the method to draw out objects and to trace the trajectory of the moving object using a CCD camera and it describes the method to recognize the shape of objects by neural network. We can transform pixel points to objects position of the real space using the proposed viewport. This proposed technique is used by the single vision system based on floor map.

• Journal of Korea Multimedia Society
• /
• v.17 no.10
• /
• pp.1150-1159
• /
• 2014

This paper proposes a method for extracting tissue cells from an organism image by an electron microscope and getting the whole cell number and the area from the cell. In general, the difference between the cell color and the background is used to extract tissue cell. However, there may be a problem when overlapped cells are seen as a single cell. To solve the problem, we split them by using cell size and curvature. This method has a 99% accuracy rate. To measure the cell area, we compute two areas, the inside and boundary of the cell. The inside is simply calculated by the number of pixels. The cell boundary is obtained by applying super sampling, linear interpolation, and cubic spline interpolation. It improves the error rate, 18%, 19%, and 120% respectively, in comparison to the counting method that counts a pixel area as 1.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.522-524
• /
• 2019

In spite of a rapid development in the quality of built-in mobile cameras, their some physical restrictions hinder them to achieve the satisfactory results of digital single lens reflex (DSLR) cameras. In this work we propose an end-to-end deep learning method to translate ordinary images by mobile cameras into DSLR-quality photos. The method is based on the framework of generative adversarial networks (GANs) with several improvements. First, we combined the U-Net with DenseNet and connected dense block (DB) in terms of U-Net. The Dense U-Net acts as the generator in our GAN model. Then, we improved the perceptual loss by using the VGG features and pixel-wise content, which could provide stronger supervision for contrast enhancement and texture recovery.

• Journal of the Semiconductor & Display Technology
• /
• v.18 no.2
• /
• pp.63-66
• /
• 2019

Due to the development of medical imaging technology, different imaging technologies provide a large amount of effective information. However, different imaging method caused the limitations of information integrity by using single type of image. Combining different image together so that doctor can obtain the information from medical image comprehensively. Image registration algorithm based on mutual information has become one of the hotspots in the field of image registration with its high registration accuracy and wide applicability. Because the information theory-based registration technology is not dependent on the gray value difference of the image, and it is very suitable for multimodal medical image registration. However, the method based on mutual information has a robustness problem. The essential reason is that the mutual information itself is not have enough information between the pixel pairs, so that the mutual information is unstable during the registration process. A large number of local extreme values are generated, which finally cause mismatch. In order to overcome the shortages of mutual information registration method, this paper proposes a registration method combined with image spatial structure information and mutual information.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.23 no.2
• /
• pp.206-216
• /
• 1986

In this paper, the moving distance and velocity of a single moving object are measured by sampling three frames in a two-dimensional line sequence image. The brightness of each frame is analyzed, and the bit data of their pixel are rearranged so that the difference image may be extracted. The parameters for recognition of the object are the gray level of the object, the number of vertex points and the distance between the vertex points. The moving distance obtained from the coordinate which is constructed by the bit processing of the data in the memory map of a microcomputer, and the moving velocity is obtained from the moving distance and the time interval between the first and second sampled frames.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.6 s.312
• /
• pp.92-100
• /
• 2006

In imaging devices such as digital cameras using a single image sensor, captured images are the sub-sampled images comprised of the pixels that have only one of the three primary colors per a pixel. This images should be restored to the color images through an image processing referred as color interpolation. In this paper, we derive relation between the average of the data from CFA image sensor and the average of each color channel data. By using this relation, a new efficient method for color interpolation is proposed. Also, in order to reduce the zipper effect in a restored image, missing luminance values are interpolated along any edges in the captured image. On the other hand, for the chrominance channel interpolation, we average difference between a chrominance value and a luminance value in a local area, and this average value is added to the pixel value of the interpolated location. The proposed method has been compared with several previous methods, and our experimental results show the better results than the other methods.

• Spatial Information Research
• /
• v.18 no.3
• /
• pp.73-83
• /
• 2010

This study presents a method to segment an image, employing the statistics observed at each pixel location across sequential frame images. In the acquisition and analysis of spatial information, utilization of digital image processing technique has very important implications. Various image segmentation techniques have been presented to distinguish the area of digital images. In this study, based on the analysis of the spectroscopic characteristics of sequential frame images that had been previously researched, an image segmentation method was proposed by using the randomness occurring among a sequence of frame images for a same scene. First of all, we computed the mean and standard deviation values at each pixel and found reliable pixels to determine seed points using their standard deviation value. For segmenting an image into individual regions, we conducted region growing based on a T-test between reference and candidate sample sets. A comparative analysis was conducted to assure the performance of the proposed method with reference to a previous method. From a set of experimental results, it is confirmed that the proposed method using a sequence of frame images segments a scene better than a method using a single frame image.

• The KIPS Transactions:PartB
• /
• v.17B no.2
• /
• pp.115-124
• /
• 2010

Stereo matching techniques are categorized in two major schemes, local and global matching techniques. In global matching schemes, several investigations are introduced, where cost accumulation is performed in multiple matching lines. In this paper, we introduce a new multi-line stereo matching techniques which expands a conventional single-line matching scheme to multiple one. Matching cost is based on simple normalized cross correlation. We expand the scan-line optimization technique to a multi-line scan-line optimization technique. The proposed technique first generates a reliability image, which is iteratively updated based on the previous reliability measure. After some number of iterations, the reliability image is completed by a hole-filling algorithm. The hole-filling algorithm introduces a disparity score table which records the disparity score of the current pixel. The disparity of an empty pixel is determined by comparing the scores of the neighboring pixels. The proposed technique is tested using the Middlebury and CMU stereo images. The error analysis shows that the proposed matching technique yields better performance than using conventional global matching algorithm.

• The Journal of the Korea Contents Association
• /
• v.9 no.8
• /
• pp.91-98
• /
• 2009

This paper presents a solution to error avalanche of deciphering where radio noise brings random bit errors in encrypted image data under ubiquitous environment. The image capturing module is to be made comprising data compression and encryption features to reduce data traffic volume and to protect privacy. Block cipher algorithms may experience error avalanche: multiple pixel defects due to single bit error in an encrypted message. The new fault tolerant scheme addresses error avalanche effect exploiting a three-dimensional data shuffling process, which disperses error bits on many frames resulting in sparsely isolated errors. Averaging or majority voting with neighboring pixels can tolerate prominent pixel defects without increase in data volume due to error correction. This scheme has 33% lower data traffic load with respect to the conventional Hamming code based approach.