• Journal of Sensor Science and Technology
• /
• v.13 no.2
• /
• pp.144-151
• /
• 2004

In this paper, we propose an active stereo surveillance system with human-like convergence function. The proposed system uses a bottom-up saliency map model with the human-like selective attention visual function to select an interesting region in each camera. and this system compares the landmarks whether the selective region in each camera finds a same region. If the left and right cameras successfully find a same landmarks, the implemented vision system focuses on the landmark. Using the motor encoder information, we can automatically obtain the depth information and resultantly construct a depth map using the depth information. Computer simulation and experimental results show that the proposed convergence method is very effective to implement the active stereo surveillance system.

• Journal of Information Processing Systems
• /
• v.9 no.4
• /
• pp.592-601
• /
• 2013

Object segmentation is a challenging task in image processing and computer vision. In this paper, we present a visual attention based segmentation method to segment small sized interesting objects in natural images. Different from the traditional methods, we first search the region of interest by using our novel saliency-based method, which is mainly based on band-pass filtering, to obtain the appropriate frequency. Secondly, we applied the Gaussian Mixture Model (GMM) to locate the object region. By incorporating the visual attention analysis into object segmentation, our proposed approach is able to narrow the search region for object segmentation, so that the accuracy is increased and the computational complexity is reduced. The experimental results indicate that our proposed approach is efficient for object segmentation in natural images, especially for small objects. Our proposed method significantly outperforms traditional GMM based segmentation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.9
• /
• pp.2458-2482
• /
• 2023

Medical image segmentation techniques based on convolution neural networks indulge in feature extraction triggering redundancy of parameters and unsatisfactory target localization, which outcomes in less accurate segmentation results to assist doctors in diagnosis. In this paper, we propose a multi-level semantic-rich encoding-decoding network, which consists of a Pooling-Conv-Former (PCFormer) module and a Cbam-Dilated-Transformer (CDT) module. In the PCFormer module, it is used to tackle the issue of parameter explosion in the conservative transformer and to compensate for the feature loss in the down-sampling process. In the CDT module, the Cbam attention module is adopted to highlight the feature regions by blending the intersection of attention mechanisms implicitly, and the Dilated convolution-Concat (DCC) module is designed as a parallel concatenation of multiple atrous convolution blocks to display the expanded perceptual field explicitly. In addition, MultiHead Attention-DwConv-Transformer (MDTransformer) module is utilized to evidently distinguish the target region from the background region. Extensive experiments on medical image segmentation from Glas, SIIM-ACR, ISIC and LGG demonstrated that our proposed network outperforms existing advanced methods in terms of both objective evaluation and subjective visual performance.

• Journal of KIISE:Software and Applications
• /
• v.31 no.2
• /
• pp.189-202
• /
• 2004

In this paper, we propose a new salient region detection method in an image. The algorithm is based on the characteristics of human's bottom-up visual attention. Several features known to influence human visual attention like color, intensity and etc. are extracted from the each regions of an image. These features are then converted to importance values for each region using its local competition function and are combined to produce a saliency map, which represents the saliency at every location in the image by a scalar quantity, and guides the selection of attended locations, based on the spatial distribution of saliency region of the image in relation to its Perceptual importance. Results shown indicate that the calculated Saliency Maps correlate well with human perception of visually important regions.

• Journal of KIISE:Software and Applications
• /
• v.26 no.10
• /
• pp.1193-1203
• /
• 1999

본 논문에서는 한글과 영문 문서 영상들에 대한 기울어짐 추정(skew estimation) 알고리즘을 제안한다. 제안 방법은 전체 문서 영상에서 텍스트 요소들이 밀집되어 있는 영역을 선별하고, 선별된 영역에 대해 허프 변환을 적용하는 선택적 주의집중(selective attention) 방식을 채택한다. 제안 방법의 기울기 추정 과정은 2단계로 구성되는데, coarse 단계에서는 전체 영상을 몇 개의 영역으로 나누고 동일한 영역에 속하는 데이타들간의 연결 각도를 계산하여 각 영역별 accumulator에 저장한다. accumulator에 저장된 빈도치를 기준으로 $\pm$45$^{\circ}$범위 내에서 최대 $\pm$1$^{\circ}$의 오차를 가진 각 영역별 기울기를 계산한 후, 이들 중 최대 빈도값을 갖는 영역을 선정하고 그 영역의 기울기 각도를 문서 영상의 대략적인 기울기 각도로 결정한다. Refine 단계에서는 coarse 단계에서 선정된 영역에 허프 변환을 적용하여 정확한 기울기를 계산하는데, coarse 단계에서 추정한 기울기의 $\pm$1$^{\circ}$범위 내에서 0.1$^{\circ}$간격으로 측정한다. 이와 같은 선택적 주의집중 방식을 통해 기울기 추정에 소요되는 시간 비용은 최소화하고, 추정의 정확도는 최대화 할 수 있다.제안 방법의 성능 평가를 위한 실험은 다양한 형태의 영문과 한글 문서 영상 2,016개에 적용되었다. 제안 방법의 평균 수행 시간은 Pentium 200MHz PC에서 0.19초이고 평균 오차는 $\pm$0.08$^{\circ}$이다. 또한 기존의 기울기 추정 방법과 제안 방법의 성능을 비교하여 제안 방법의 우수성을 입증하였다.Abstract In this paper we propose a skew estimation algorithm for English and Korean document images. The proposed method adopts a selective attention strategy, in which we choose a region of interest which contains a cluster of text components and then apply a Hough transform to this region. The skew estimation process consists of two steps. In the coarse step, we divide the entire image into several regions, and compute the skew angle of each region by accumulating the slopes of lines connecting any two components in the region. The skew angle is estimated within the range of $\pm$45 degree with a maximum error of $\pm$1 degree. Next we select a region which has the most frequent slope in the accumulators and determine the skew angle of the image roughly as the angle corresponding to the most frequent slope. In the refine step, a Hough transform is applied for the selected region within the range of $\pm$1 degree along the angle computed from the coarse step, with an angular resolution of 0.1 degree. Based on this selective attention strategy, we can minimize the time cost and maximize the accuracy of the skew estimation.We have measured the performance of the proposed method by an experiment with 2,016 images of various English and Korean documents. The average run time is 0.19 second on a Pentium 200MHz PC, and the average error is $\pm$0.08 degree. We also have proven the superiority of our algorithm by comparing the performance with that of other well-known methods in the literature.

• The KIPS Transactions:PartB
• /
• v.15B no.3
• /
• pp.189-196
• /
• 2008

In this paper, a CMOS digital image sensor, which consists of A/D conversion, motion estimation circuits, and an attention module for ROI (Region of Interest) detection is presented. The functions of A/D conversion and motion estimation are implemented by $0.6{\mu}m$ CMOS processing circuit as hardware, and the attention module is implemented outside the circuit as software currently. Attention modules are taken to improve limited applications of the smart image sensor. The current smart image sensor responses to the changes of intensity, and uses the integration time to estimate motion. Therefore it is limited in its applications. To make up for inherent property of the sensor from circuit design and extend its applications we decide to introduce perception solutions to the image sensor. Attention modules for still and moving images are employed to achieve such purposes. The suggested approach makes the smart image sensor available with additional functions for such cases that motion estimation or intensity changes are not observed. Experimental result shows the usefulness and extension of the image sensor.

• Proceedings of the Korea Concrete Institute Conference
• /
• 2002.05a
• /
• pp.291-296
• /
• 2002

In order to apply precast decks to the continuous composite bridges, several experiments and analytical studies were performed. For the continuous composite bridges, special attention should be paid to the transverse joints in negative moment region. Judging from the results, combination of longitudinal internal prestressing tendon and the external tendon can be effectively used for the prevention of cracking in the negative moment region of precast decks.

• Proceedings of the Plant Resources Society of Korea Conference
• /
• 2003.10a
• /
• pp.19-19
• /
• 2003

The region activation have attracted much attention of politician, government, scholar, local residents and so on, but it is very difficult and delicate. The policy of Inje-Gun government are jump into the lucrative tourism, the strong agriculture, the comfortable welfare and the attentive administration. But the present status is very poor. Economic self-supporting index is only 19%. The policy will achieve by increase of economic self-supporting. I am going to propose some item for region activation in Inje-Gun. The items are medicinal garden construction, green tourism, local festival, residents increase and so on.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.4
• /
• pp.943-950
• /
• 2011

In this paper, we presented an attention measurement method in 2D and 3D image/video to be applied for image and video retargeting and compressive processing. 2D attention is derived from the three main components, intensity, color, and orientation, while depth information is added for 3D attention. A rarity-based attention method is presented to obtain more interested region or objects. Displaced depth information is matched to attention probability in distorted stereo images and finally a stereo distortion predictor is designed by integrating low-level HVS responses. As results, more efficient attention scheme is developed from the conventional methods and performance is proved by applying for video retargeting.

• Journal of Korea Multimedia Society
• /
• v.24 no.10
• /
• pp.1380-1390
• /
• 2021

The video face recognition (FR) is one of the most popular researches in the field of computer vision due to a variety of applications. In particular, research using the attention mechanism is being actively conducted. In video face recognition, attention represents where to focus on by using the input value of the whole or a specific region, or which frame to focus on when there are many frames. In this paper, we propose a novel attention based deep learning method. Main novelties of our method are (1) the use of combining two loss functions, namely weighted Softmax loss function and a Triplet loss function and (2) the feasibility of end-to-end learning which includes the feature embedding network and attention weight computation. The feature embedding network has a positive effect on the attention weight computation by using combined loss function and end-to-end learning. To demonstrate the effectiveness of our proposed method, extensive and comparative experiments have been carried out to evaluate our method on IJB-A dataset with their standard evaluation protocols. Our proposed method represented better or comparable recognition rate compared to other state-of-the-art video FR methods.