• Journal of Broadcast Engineering
• /
• v.13 no.2
• /
• pp.214-221
• /
• 2008

In this paper, we proposed a new mesh editing algorithm based on the motion between two sample meshes. First, the motion vectors are defined as the derivation vector of the corresponding vertices on the sample meshes. Then, the motion feature vectors are extracted between the motion vectors. The motion feature vectors represent the similarity of the vertex motion in a local mesh surface. When a mesh structure is forced by an external motion of anchor vertices, the deformed mesh geometry is obtained by minimizing the cost function with preserving the motion feature vectors. Simulation results demonstrated that the proposed algorithm yields visually pleasing editing results.

• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.47-60
• /
• 2012

Video data comes in the form of the unstructured and the complex structure. As the importance of efficient management and retrieval for video data increases, studies on the video parsing based on the visual features contained in the video contents are researched to reconstruct video data as the meaningful structure. The early studies on video parsing are focused on splitting video data into shots, but detecting the shot boundary defined with the physical boundary does not cosider the semantic association of video data. Recently, studies on structuralizing video shots having the semantic association to the video scene defined with the semantic boundary by utilizing clustering methods are actively progressed. Previous studies on detecting the video scene try to detect video scenes by utilizing clustering algorithms based on the similarity measure between video shots mainly depended on color features. However, the correct identification of a video shot or scene and the detection of the gradual transitions such as dissolve, fade and wipe are difficult because color features of video data contain a noise and are abruptly changed due to the intervention of an unexpected object. In this paper, to solve these problems, we propose the Scene Detector by using Color histogram, corner Edge and Object color histogram (SDCEO) that clusters similar shots organizing same event based on visual features including the color histogram, the corner edge and the object color histogram to detect video scenes. The SDCEO is worthy of notice in a sense that it uses the edge feature with the color feature, and as a result, it effectively detects the gradual transitions as well as the abrupt transitions. The SDCEO consists of the Shot Bound Identifier and the Video Scene Detector. The Shot Bound Identifier is comprised of the Color Histogram Analysis step and the Corner Edge Analysis step. In the Color Histogram Analysis step, SDCEO uses the color histogram feature to organizing shot boundaries. The color histogram, recording the percentage of each quantized color among all pixels in a frame, are chosen for their good performance, as also reported in other work of content-based image and video analysis. To organize shot boundaries, SDCEO joins associated sequential frames into shot boundaries by measuring the similarity of the color histogram between frames. In the Corner Edge Analysis step, SDCEO identifies the final shot boundaries by using the corner edge feature. SDCEO detect associated shot boundaries comparing the corner edge feature between the last frame of previous shot boundary and the first frame of next shot boundary. In the Key-frame Extraction step, SDCEO compares each frame with all frames and measures the similarity by using histogram euclidean distance, and then select the frame the most similar with all frames contained in same shot boundary as the key-frame. Video Scene Detector clusters associated shots organizing same event by utilizing the hierarchical agglomerative clustering method based on the visual features including the color histogram and the object color histogram. After detecting video scenes, SDCEO organizes final video scene by repetitive clustering until the simiarity distance between shot boundaries less than the threshold h. In this paper, we construct the prototype of SDCEO and experiments are carried out with the baseline data that are manually constructed, and the experimental results that the precision of shot boundary detection is 93.3% and the precision of video scene detection is 83.3% are satisfactory.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.59-68
• /
• 2022

This paper presents a method for 1:1 verification by comparing the similarity between the given real product image and the drawing image. The proposed method combines two existing CNN-based deep learning models to construct a Siamese Network. After extracting the feature vector of the image through the FC (Fully Connected) Layer of each network and comparing the similarity, if the real product image and the drawing image (front view, left and right side view, top view, etc) are the same product, the similarity is set to 1 for learning and, if it is a different product, the similarity is set to 0. The test (inference) model is a deep learning model that queries the real product image and the drawing image in pairs to determine whether the pair is the same product or not. In the proposed model, through a comparison of the similarity between the real product image and the drawing image, if the similarity is greater than or equal to a threshold value (Threshold: 0.5), it is determined that the product is the same, and if it is less than or equal to, it is determined that the product is a different product. The proposed model showed an accuracy of about 71.8% for a query to a product (positive: positive) with the same drawing as the real product, and an accuracy of about 83.1% for a query to a different product (positive: negative). In the future, we plan to conduct a study to improve the matching accuracy between the real product image and the drawing image by combining the parameter optimization study with the proposed model and adding processes such as data purification.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.2
• /
• pp.117-128
• /
• 2001

In order to achieve optimal retrieval, i.e., to find the best match to a query according to a certain similarity measure, the exhaustive search should be performed literally for all the images in a database. However, the straightforward exhaustive search algorithm is computationally expensive in large image databases. To reduce its heavy computational cost, this paper presents a fast exhaustive multi-resolution search algorithm based on image database clustering. Firstly, the proposed algorithm partitions the whole image data set into a pre-defined number of clusters having similar feature contents. Next, for a given query, it checks the lower bound of distances in each cluster, eliminating disqualified clusters. Then, it only examines the candidates in the remaining clusters. To alleviate unnecessary feature matching operations in the search procedure, the distance inequality property is employed based on a multi-resolution data structure. The proposed algorithm realizes a fast exhaustive multi-resolution search for either the best match or multiple best matches to the query. Using luminance histograms as a feature, we prove that the proposed algorithm guarantees optimal retrieval with high searching speed.

• The KIPS Transactions:PartB
• /
• v.19B no.1
• /
• pp.1-8
• /
• 2012

Mosaic is to make a big image by gathering lots of small materials having various colors. With advance of digital imaging techniques, photomosaic techniques using photos are widely used. In this paper, we presents an automatic photomosaic algorithm based on adaptive tiling and block matching. The proposed algorithm is composed of two processes: photo database generation and photomosaic generation. Photo database is a set of photos (or tiles) used for mosaic, where a tile is divided into $4{\times}4$ regions and the average RGB value of each region is the feature of the tile. Photomosaic generation is composed of 4 steps: feature extraction, adaptive tiling, block matching, and intensity adjustment. In feature extraction, the feature of each block is calculated after the image is splitted into the preset size of blocks. In adaptive tiling, the blocks having similar similarities are merged. Then, the blocks are compared with tiles in photo database by comparing euclidean distance as a similarity measure in block matching. Finally, in intensity adjustment, the intensity of the matched tile is replaced as that of the block to increase the similarity between the tile and the block. Also, a tile redundancy minimization scheme of adjacent blocks is applied to enhance the quality of mosaic photos. In comparison with Andrea mosaic software, the proposed algorithm outperforms in quantitative and qualitative analysis.

• Korean Journal of Remote Sensing
• /
• v.37 no.3
• /
• pp.475-491
• /
• 2021

With the recent increase in available high-resolution (< ~1 m) satellite SAR images, the demand for precise registration of SAR images is increasing in various fields including change detection. The registration between high-resolution SAR images acquired in different look angle is difficult due to speckle noise and geometric distortion caused by the characteristics of SAR images. In this study, registration is performed in two stages, coarse and fine, using the x-band SAR data imaged at staring spotlight mode of TerraSAR-X. For the coarse registration, a method combining the adaptive sampling method and SAR-SIFT (Scale Invariant Feature Transform) is applied, and three rigid methods (NCC: Normalized Cross Correlation, Phase Congruency-NCC, MI: Mutual Information) and one non-rigid (Gefolki: Geoscience extended Flow Optical Flow Lucas-Kanade Iterative), for the fine registration stage, was performed for performance comparison. The results were compared by using RMSE (Root Mean Square Error) and FSIM (Feature Similarity) index, and all rigid models showed poor results in all image combinations. It is confirmed that the rigid models have a large registration error in the rugged terrain area. As a result of applying the Gefolki algorithm, it was confirmed that the RMSE of Gefolki showed the best result as a 1~3 pixels, and the FSIM index also obtained a higher value than 0.02~0.03 compared to other rigid methods. It was confirmed that the mis-registration due to terrain effect could be sufficiently reduced by the Gefolki algorithm.

• The KIPS Transactions:PartB
• /
• v.17B no.1
• /
• pp.37-46
• /
• 2010

In this paper we have presented a two dimensional model based tracking system using improved chamfer matching. Conventional chamfer matching could not calculate similarity well between the object and image when there is very cluttered background. Then we have improved chamfer matching to calculate similarity well even in very cluttered background with edge and corner feature points. Improved chamfer matching is used as likelihood function of particle filter which tracks the geometric object. Geometric model which uses edge and corner feature points, is a discriminant descriptor in color changes. Particle Filter is more non-linear tracking system than Kalman Filter. Then the presented method uses geometric model, particle filter and improved chamfer matching for tracking object in complex environment. In experimental result, the robustness of our system is proved by comparing other methods.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.247-255
• /
• 2015

This paper presents a study on the performance of the music search based on the automatically recognized music-emotion labels. As in the other media data, such as speech, image, and video, a song can evoke certain emotions to the listeners. When people look for songs to listen, the emotions, evoked by songs, could be important points to consider. However; very little study has been done on the performance of the music-emotion labels to the music search. In this paper, we utilize the three axes of human music perception (valence, activity, tension) and the five basic emotion labels (happiness, sadness, tenderness, anger, fear) in measuring music similarity for music search. Experiments were conducted on both genre and singer datasets. The search accuracy of the proposed emotion-based music search was up to 75 % of that of the conventional feature-based music search. By combining the proposed emotion-based method with the feature-based method, we achieved up to 14 % improvement of search accuracy.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.7
• /
• pp.816-822
• /
• 2016

Content-Based image retrieval is a method to search by image features such as local color, texture, and other image content information, which is different from conventional tag or labeled text-based searching. In real life data, the number of images having tags or labels is relatively small, so it is hard to search the relevant images with text-based approach. Existing image search method only based on image feature similarity has limited performance and does not ensure that the results are what the user expected. In this study, we propose and validate a machine learning based approach to improve the performance of the image search engine. We note that when users search relevant images with a query image, they would expect the retrieved images belong to the same category as that of the query. Image classification method is combined with the traditional image feature similarity method. The proposed method is extensively validated on a public PASCAL VOC dataset consisting of 11,530 images from 20 categories.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.1
• /
• pp.20-27
• /
• 2003

In this paper, we implemented a content-based image retrieval system that user can choose a wanted query region of object and retrieve similar object from image database. Query image is induced to wavelet transformation after divided into hue components and gray components that hue features is extracted through color autocorrelogram and dispersion in hue components. Texture feature is extracted through autocorrelogram and GLCM in gray components also. Using features of two components, retrieval is processed to compare each similarity with database image. In here, weight value is applied to each similarity value. We make up for each defect by deriving features from two components beside one that elevations of recall and precision are verified in experiment results. Moreover, retrieval efficiency is improved by weight value. And various features of database images are indexed automatically in feature library that make possible to rapid image retrieval.