• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.28 no.2
• /
• pp.125-132
• /
• 2004

In this paper, we propose a new two-stage segmentation method for the effective object recognition which uses region-growing algorithm and k-means clustering method. At first, an image is segmented into many small regions via region growing algorithm. And then the segmented small regions are merged in several regions so that the regions of an object may be included in the same region using typical k-means clustering method. This paper also establishes similarity measurement which is useful for object recognition in an image. Similarity is measured by fuzzy system whose input variables are compactness, magnitude of biasness and orientation of biasness of the object image, which are geometrical features of the object. To verify the effectiveness of the proposed two-stage segmentation method and similarity measurement, experiments for object recognition were made and the results show that they are applicable to object recognition under normal circumstance as well as under abnormal circumstance of being.

• Journal of information and communication convergence engineering
• /
• v.19 no.2
• /
• pp.79-83
• /
• 2021

This paper proposes a novel image classification method based on few-shot learning, which is mainly used to solve model overfitting and non-convergence in image classification tasks of small datasets and improve the accuracy of classification. This method uses model structure optimization to extend the basic convolutional neural network (CNN) model and extracts more image features by adding convolutional layers, thereby improving the classification accuracy. We incorporated certain measures to improve the performance of the model. First, we used general methods such as setting a lower learning rate and shuffling to promote the rapid convergence of the model. Second, we used the data expansion technology to preprocess small datasets to increase the number of training data sets and suppress over-fitting. We applied the model to 10 monkey species and achieved outstanding performances. Experiments indicated that our proposed method achieved an accuracy of 87.92%, which is 26.1% higher than that of the traditional CNN method and 1.1% higher than that of the deep convolutional neural network ResNet50.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.12
• /
• pp.3364-3382
• /
• 2023

Remote sensing image segmentation plays an important role in realizing intelligent city construction. The current mainstream segmentation networks effectively improve the segmentation effect of remote sensing images by deeply mining the rich texture and semantic features of images. But there are still some problems such as rough results of small target region segmentation and poor edge contour segmentation. To overcome these three challenges, we propose an improved semantic segmentation model, referred to as MRU-Net, which adopts the U-Net architecture as its backbone. Firstly, the convolutional layer is replaced by BasicBlock structure in U-Net network to extract features, then the activation function is replaced to reduce the computational load of model in the network. Secondly, a hybrid multi-scale recognition module is added in the encoder to improve the accuracy of image segmentation of small targets and edge parts. Finally, test on Massachusetts Buildings Dataset and WHU Dataset the experimental results show that compared with the original network the ACC, mIoU and F1 value are improved, and the imposed network shows good robustness and portability in different datasets.

• Proceedings of the IEEK Conference
• /
• 2002.06b
• /
• pp.83-86
• /
• 2002

We suggest binary image decryption system using two-wavc mixing in Photorefractive crystal. Compared with a conventional method, this method can make optical alignment easily, and brighten the encrypted image even if a small input signal, by index grating of photorefractive crystal. Also it can reconstruct the encrypted image by only reference beam in real time.

• ETRI Journal
• /
• v.28 no.6
• /
• pp.819-821
• /
• 2006

Log-polar coordinate image space is proposed as a solution for the problem of unbounded accumulator space in the automatic detection of vanishing points. The proposed method can detect vanishing points at high speed under small memory requirements, as opposed to conventional image space based methods.

• Proceedings of the Korean Society of Agricultural Engineers Conference
• /
• 1999.10c
• /
• pp.519-524
• /
• 1999

In order to assess the estimtions of CN for a small agricultural watershed using the satellite image, TM image from Landsat-5 was classsified by MLC. CN for each pixels in the image was estimaed using the results. For the estimation enhancing , it was tried that each land use area in a pixel was estimated by the mixel assumption and the averaged CN by weight areas. Those resutls were applied for the actual hydrologic analyses were highly concerned with the observed runoff discharge and more enhanced on the mixel assumption.

• Proceedings of the IEEK Conference
• /
• 2002.06a
• /
• pp.207-210
• /
• 2002

We suggest binary image decryption system using two-wave mixing in photorefractive crystal. Compared with a conventional method, this method can make optical alignment easily, and brighten the encrypted image even if a small input signal, by index grating of photorefractive crystal. Also it can reconstruct the encrypted image by only reference beam in real time.

• Proceedings of the KIEE Conference
• /
• 1999.11c
• /
• pp.781-783
• /
• 1999

A storage or transmission of image is difficult without image compression processing because the numbers of generated or reborned image data are very much. In case of the random signal, image compression efficiency is low doing without loss of image information, but compressibility by using JPEG is better. We used Huffman code of JPEG, it assigne the low bit value for data of a lot of generated frequency, assigne the high bit value for data of a small quantity. This paper improved image compression efficiency with transformming Q-factor and certified the results with compressed image. A proposed method is very efficience for continuos a color tone image.

• Proceedings of the KIEE Conference
• /
• 1996.07b
• /
• pp.1368-1370
• /
• 1996

The Information of Edges which takes small area comparing with the total image is very important in DNA images as well as general images. DNA image is the object should be managed by computing and it's demanding information is less than general images, but the accuracy is more important In a huge DNA image processing system such as DNA Information Bank, the performance depends on the size of image. In this paper, we extract the edge information and make it as a binary image. To reduce the size of the original image, it was applied by JPEG image compression method. The compressed image is combined with edge information when they are restored. As a result, Both the image compression ratio and restoration quality are optimized without the loss of critical information.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1706-1725
• /
• 2024

The unique U-shaped structure of U-Net network makes it achieve good performance in image segmentation. This network is a lightweight network with a small number of parameters for small image segmentation datasets. However, when the medical image to be segmented contains a lot of detailed information, the segmentation results cannot fully meet the actual requirements. In order to achieve higher accuracy of medical image segmentation, a novel improved U-Net network architecture called multi-scale encoder-decoder U-Net+ (MEDU-Net+) is proposed in this paper. We design the GoogLeNet for achieving more information at the encoder of the proposed MEDU-Net+, and present the multi-scale feature extraction for fusing semantic information of different scales in the encoder and decoder. Meanwhile, we also introduce the layer-by-layer skip connection to connect the information of each layer, so that there is no need to encode the last layer and return the information. The proposed MEDU-Net+ divides the unknown depth network into each part of deconvolution layer to replace the direct connection of the encoder and decoder in U-Net. In addition, a new combined loss function is proposed to extract more edge information by combining the advantages of the generalized dice and the focal loss functions. Finally, we validate our proposed MEDU-Net+ MEDU-Net+ and other classic medical image segmentation networks on three medical image datasets. The experimental results show that our proposed MEDU-Net+ has prominent superior performance compared with other medical image segmentation networks.