F. Arman, A. Hsu and M.Y. Chiu, "Image processing on compressed data for large video databases", Proc. ACM Multimedia 93, pp. 267-272, Anaheim, CA, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....data, which must be restored from the compressed data by the time consuming decompression operation. Since most digital videos are stored in a compressed format such as MPEG, several algorithms for detecting abrupt shot boundaries directly in the compressed domain have recently emerged [1,4,7,11]. These methods use the information directly available from an MPEG stream, such as DCT coefficients, motion vectors and bit rates, as dissimilarity measures. Since decompression is not necessary, these compressed domain methods are computationally more efficient. In addition, some of these ....

F. Arman, A. Hsu and M.Y. Chiu, "Image processing on compressed data for large video databases", in Proceedings of the 2nd ACM International Conference on Multimedia, Anaheim, California, August 1993, pp. 267-272.

....mosaic generation techniques, the sprite generation is a lossy process. Sprite generation is closely related with extraction of good features. These features may also need to be extracted in the compressed domain. Feature extraction and edge detection in compressed DCT domain have been studied in [70, 3, 85]. Patterns of the first AC coefficients by zigzag ordering have been used for coarse edge detection [85] The number of non zero coefficients has been studied in [3] First two AC coefficients are used for edge and shape detection in [70] The general problem with the coarse edge detection ....

....to be extracted in the compressed domain. Feature extraction and edge detection in compressed DCT domain have been studied in [70, 3, 85] Patterns of the first AC coefficients by zigzag ordering have been used for coarse edge detection [85] The number of non zero coefficients has been studied in [3]. First two AC coefficients are used for edge and shape detection in [70] The general problem with the coarse edge detection problems is that most of the weight is given to coefficients that appear first according to the zigzag ordering. In traditional mosaicing methods, mosaics are generated by ....

[Article contains additional citation context not shown here]

F. Arman and et al. Image processing on compressed data for large video databases. In Proc. ACM Int. Conf. Multimedia, June 1993.

....with these di#erence based algorithms is that they are sensitive to busy scenes, in which intensities change substantially from frame to frame due to camera object motion. Since the availability of MPEG video, several algorithms for detecting cuts directly in the MPEG domain have emerged [AHC93, YL95, SD95, ZLS95, LZ95, FLM96, KDR96, IP97, GHP98, JHEJ98, KKC99, MIP99] These methods use the information directly available from an MPEG stream, such as DCT coe#cients, motion vectors and bit rates, to calculate the frame dissimilarity. A full review of cut detection algorithms operating ....

F. Arman, A. Hsu, and M-Y. Chiu. Image processing on compressed data for large video databases. In Proc. of ACM multimedia '93, pages 267--272, 1993.

....for a certain span of time is considered , unsent . Since WebMovie composes the layers of the respective stream dynamically, lost data is automatically retransmitred in a later iteration of the transmission (see figure 6) data loss Information atom retransmission : 1418 1121171 1251291321351391q31 n l t Iteration n 2 Iteration figure 6: principle of delayed data retransmission After the transmission has been completed all the data has been delivered. This elegant integration of error correction Into the iterative movie transmission comes at no extra cost to the ....

....for a certain span of time is considered , unsent . Since WebMovie composes the layers of the respective stream dynamically, lost data is automatically retransmitred in a later iteration of the transmission (see figure 6) data loss Information atom retransmission : 1418 1121171 1251291321351391q31 n l t Iteration n 2 Iteration figure 6: principle of delayed data retransmission After the transmission has been completed all the data has been delivered. This elegant integration of error correction Into the iterative movie transmission comes at no extra cost to the ....

[Article contains additional citation context not shown here]

Arman, F., Hsu, A., Chiu, M. ,,Image Processing on Compressed Data for Large Video Databases". Proceedings of the I st ACM International Confrererice on Multimedia. August 1993.

....or partial decoding) is needed to obtain DCT coefficients in P and 16 B frames. Note that DCT coefficients of low intensity blocks are prone to coding noise, which should be considered when using them for indexing or matching. a. Correlation between DCT coefficients of two frames Arman et al. [4] use a subset of DCT coefficients of a subset of blocks to form feature vectors for each frame. Then the angles between the feature vectors of different frames are used to measure their similarity. They use this method to detect shot boundaries and show its effectiveness for cut detection. The ....

F. Arman, A. Hsu, and M.-Y. Chiu, "Image processing on compressed data for large video databases," Proc. ACM Multimedia 93, pp. 267-272, 1993, Anaheim, CA.

....of the viewer. We believe these new functions are the key to make MPI video a revolutionary new media. To make such functions possible, however, much advancement of technology is required in the fields of computer vision, multimedia databases and human interface design, see [20, 18, 10, 9] and [21, 22, 15, 2, 4, 16, 1]. 6 Conclusion and Future Research In this paper, we presented our ideas of infoscopes. Infoscopes provide microscopes as well as telescopes for the information systems of the future. These systems use databases for multimedia data, and provide very rich semantics and techniques to recover ....

Farshid Arman, Arding Hsu, and Ming-Yee Chiu. Image processing on compressed data for large video databases. In Proceedings of the ACM MultiMedia, pages 267--272, California, USA, June 1993. Association of Computing Machinery.

....that limit the robustness of these algorithms. One of the common problems in robust shot detection results from the fact that there are many flashlights in news video, which often introduce false detection of shot boundaries. Only some simple solutions to this problem have been proposed in [2]. Their main limitations are that they assume the flashlight just occur during one frame. In real world, such as news video, there will be many flashlights occur during a period of time and influence multiple consecutive frames. Another problem that has not been solved very effectively well is ....

F. Arman, A. Hsu, and M.Y. Chiu, Image processing on compressed data for large video databases, Proceedings of 1st ACM International Conference on Multimedia, Anaheim, CA, pp.267-272, 1993.

....If one was able to analyse video directly in compressed format, one would save both the auxiliary storage for decompressed data and the computational costs for decompression. Several methods have been proposed for scene change detection on compressed MPEG [22, 32, 20, 47, 41] and Motion JPEG [4] data. These methods have proved to be sufficiently accurate for segmentation of the majority of shots in a video sequence. Before describing the approaches for MPEG data, a brief description on the MPEG standard can be useful. MPEG defines three different types of frames. I frames (intracoded ....

F. Arman, A. Hsu, M. Chiu. Image processing on compressed data for large video databases. In Proceedings of First ACM International Conference on Multimedia, pp. 267--272, August 1993.

....or partial decoding) is needed to obtain DCT coefficients in P and B frames. Note that DCT coefficients of low intensity blocks are prone to coding noise, which should be considered when using them for indexing or matching. a. Correlation between DCT coefficients of two frames Arman et al. [4] use a subset of DCT coefficients of a subset of blocks to form feature vectors for each frame. Then the angles between the feature vectors of different frames are used to measure their similarity. They use this method to detect shot boundaries and show its effectiveness for cut detection. The ....

F. Arman, A. Hsu, and M.-Y. Chiu, "Image processing on compressed data for large video databases," Proc. ACM Multimedia 93, pp. 267-272, 1993, Anaheim, CA.

....the video which allows the content based access of video. This paper presents the essential characteristics of the metadata necessary to provide content based video access. Research in the area of digital video analysis and video databases has been focussed on the fine grain manipulation of video [2, 5, 15, 19, 21]. Some current research is addressing the problem of video retrieval at a slightly coarser grain. Davenport et al. [7] and Davis [9] address the problem of manual video annotation and the types of annotation symbols needed. Hampapur [14] address the problem of video indexing based on different ....

Farshid Arman, Arding Hsu, and Ming-Yee Chiu. Image processing on compressed data for large video databases. In Proceedings of the ACM MultiMedia, pages 267--272, California, USA, June 1993. Association of Computing Machinery.

....possibilities: intra, forward, backward or interpolated. For more information about MPEG see [15] Figure 9, Figure 10 21 2.2. 2 Temporal Video Segmentation Based on DCT Coefficients The pioneering work on video parsing directly in compressed domain is conducted by Arman, Hsu and Chiu [5] who proposed a technique for cut detection based on the DCT coefficients of I frames. For each frame a subset of the DCT coefficients of a subset of the blocks is selected in order to construct a vector , 3 2 1 c c c V i . i V represents the frame i from the video sequence in the DCT ....

....It operates hierarchically by first locating a suspected cut between two I frames, then between the P frames of the GOP and finally (if necessary) by checking the B frames. 2.2. 8 Comparison of Algorithms for Temporal Video Segmentation in Compressed Domain In [12] the approaches of Arman et al. [5], Patel and Sethi [29] Meng et al. 24] Yeo and Liu [42] and Shen and Delp [33] are compared along several parameters: classification performance (recall and precision) full data use, ease of implementation, source effects. Ten MPEG video sequences containing more than 30 000 frames connected ....

F. Arman, A. Hsu, M-Y Chiu, Image processing on compressed data for large video databases, in: Proc. First ACM Intern. Conference on Multimedia, 1993, pp. 267-272.

....are available. Then, analogously to the pixel domain methods, changes between successive frames are evaluated by difference metrics and the decision is taken by complex thresholding. For example, the pioneering work on video parsing directly in compressed domain is conducted by Arman et al. [5] who have proposed a technique for cut detection based on the DCT coefficients of I frames. For each frame a subset of the DCT coefficients is selected as a representing vector and the normalized inner product between the vectors of consecutive I frames is thresholded. Yeo and Liu [33] have ....

F. Arman, A. Hsu, M-Y Chiu, "Image processing on compressed data for large video databases," in: Proc. First ACM Intern. Conference on Multimedia, , pp. 267-272, 1993.

....media and provides a unified query language and indexing structures to access the resulting integration. Hence, one such media instance we could integrate is the OVID system, though our framework is general enough to integrate many other media (which OVID cannot) In a similar vein, Arman et. al [3] develop techniques to create large video databases by processing incoming video data so as to identify features and set up access structures. Cardenas, et al. 9] have developed a query language called PICQUERY for querying certain kinds of federated multimedia systems. In contrast to their ....

F. Arman, A. Hsu and M. Chiu. (1993) Image Processing on Compressed Data for Large Video Databases, First ACM Intl. Conf. on Multimedia, pps 267--272.

....the most basic operations. To clarify this we note that most of the processing and interaction currently provided by multimedia applications is based on the raw format data. Although the provision of operations for other formats such as JPEG has been studied to some degree (Smith Rowe 1993, Arman, Hsu Chiu 1993) it is still fair to say that in most systems, the only operation available for JPEG format data is decode . Obviously, if operations could be performed on compressed format data directly, many computational savings could be made. But we may take this idea a step further by observing that the ....

Arman, F., Hsu, A. & Chiu, M. (1993), Image processing on compressed data for large video databases, in `ACM Multimedia '93'.

....algorithms [8, 14, 15, 17] that also combine spatial information with color histograms. Finally, we present some possible extensions to CCV s. COLOR HISTOGRAMS Color histograms are frequently used to compare images. Examples of their use in multimedia applications include scene break detection [1, 7, 10, 12, 22] and querying a database of images [3, 4, 11, 13] Their popularity stems from several factors. Color histograms are computationally trivial to compute. Small changes in camera viewpoint tend not to e#ect color histograms. Di#erent objects often have distinctive color histograms. ....

Farshid Arman, Arding Hsu, and MingYee Chiu. Image processing on compressed data for large video databases. In ACM Multimedia Conference, pages 267-- 272, 1993.

....the result to be computed using only shifts and adds. Shen and Sethi have similarly examined inner block transforms, operations whose range is confined to a single 8 by 8 block [20] Arman has developed a technique to detect scene changes in motion JPEG compressed video data in the DCT domain [21]. Seales has examined the problem of object recognition in the compressed domain [22] Broadhead and Owen have extended these techniques MPEG compressed audio data [23] Many extensions are possible to the work presented in this paper. Condensation algorithms can be developed that improve the ....

F. Arman, A. Hsu, M. Y. Chiu, Image Processing on Compressed Data for Large Video Databases, Proceedings of the First ACM International Conference on Multimedia, August 1993, Anaheim, Calif.

....method still requires the uncompressed mesh representation to be present in the memory. Finally, there exists a large body of research involving compression domain processing of di erent multimedia data types. This includes direct manipulation of compressed image [SS95, SR93] compressed video [AHC93, CM95] compressed text [ZM95, VC97] and compressed volume [LGK97, C 97] data. 22 5.2 Triangle Mesh Editing Operations In this subsection, we describe the set of editing operation that the proposed Compression Domain Mesh Editing (CDE) scheme supports. We believe they form a ....

F. Arman, A. Hsu, and M.Y. Chiu. Image Processing on Compressed Data for Large Video Databases. In Proceedings of the 1st ACM International Conference on Multimedia, 1993.

....proposed to further exploit the encoding strategies in the case of MPEG 1 and MPEG 2. Using the different characteristics of the three types of frames (I , P and B frames) temporal segmentation can be done at several levels of resolution and combining various features. The method presented in [3] performs first a simplification of the sequence only retaining the I frames. The feature space is some of the DCT coefficients of a selected set of blocks in the frame. For each frame, a vector is formed chaining the selected DCT coefficients. The transition is estimated by computing the ....

F. Arman, A. Hsu, and M. Y. Chiu. Image processing on compressed data for large video databases. In Proceedings 1st ACM international Conference on Multimedia, pages 267--272, August 1993.

....to refer to a sequence of video representing continuous action and story to refer to a sequence of scenes. The problem of identifying cuts (sharp transitions between shots) has been typically approached from a bottom up perspective, looking for rapid changes in color histogram or image intensity (Arman, 1993; Nagasaka, 1992; Zhang, 1993) Model based algorithms have also been developed to successfully detect fades, dissolves, and page translate edits (Hampapur, 1994) Once shots have been identified, keys frames which characterize the shot can be selected by considering the motion of objects within ....

....by proposing a model for video based on the production process and classifying video edit effects based on these models. The edit effect models are used to design feature detectors, which are used in a feature based classification approach to segment the video. Arman, Hsu, and Chui (Arman, 1993) presented a technique operating directly on compressed video detect shot boundaries. Their technique relies on the properties of the coefficients of the discrete cosine transform used in encoding the video to detect the transitions. Shot detection in the VISION system is performed by combining ....

Arman, F., Hsu, A., et al, (1993). Image Processing on Compressed Data for Large Video Databases, ACM Multimedia '93, California, USA, 267-272.

....to address image and video content. For example, it would be advantageous to search an image video database using only compressed domain operations. Several researchers have pursued this idea. Two notable contributions are Seales work on object recognition [11] and Arman s work on cut detection [2]. Seales examined searching a database of mug shots in the compressed domain using an eigenspace method. The basic idea of this work is to represent a large dimensional object (an image) in a lower dimensional subspace. The basis vectors for the subspace are images, called eigen faces , ....

Farshid Arman, Arding Hsu, Ming-Yee Chiu, Image Processing on Compressed Data for Large Video Databases, Proceedings of the First ACM International Conference on Multimedia, August 1993, Anaheim, Calif.

....Shen and Sethi [6] perform edge detection in the compressed domain, showing that a coarse edge map can be obtained up to 20 times more efficiently. In terms of video segmentation and content analysis, the problem of scene change, or cut detection, has been solved in the compressed domain [7]. There are also algorithms for resource bandwidth allocation [8] which rely on predictions using cues extracted from compressed video. The work of the Photobook project by Pentland and co workers [9,10] uses compression schemes that preserve semantic content and are thus searchable. The ....

F. Arman, A. Hsu and M.-Y. Chin, Image processing on compressed data for large video databases, in: Proceedings of the First ACM International Conference on Multimedia, August 1993.

....Shen and Sethi [16] perform edge detection in the compressed domain, showing that a coarse edgemap can be obtained up to 20 times more efficiently. In terms of video segmentation and content analysis, the problem of scene change, or cut detec2 tion, has been solved in the compressed domain [2]. There are also algorithms for resource bandwidth allocation [9] which rely on predictions using cues extracted from compressed video. The work of the Photobook project by Pentland et al. 15, 14] uses compression schemes that preserve semantic content and are thus searchable. The compression ....

F. Arman, A. Hsu, and M.-Y. Chiu. Image processing on compressed data for large video databases. In Proceedings of the First ACM International Conference on Multimedia, August 1993.

....by the use of digital video. For digital video, a significant amount of research effort has been elaborated upon probing the video content and doing the index search and pattern matching [1, 5, 14, 16] These operations were also intended to be done when the digital video is in its compressed form [2]. In view of the increasing popularity of digital video, it is important to explore the approach of content based downloading and to incorporate it into client stations and home VCR s. Consequently, we examine in this study a content based method to download record digital video from networks to ....

F. Arman, A. Hsu, and M. Y. Chiu. Image Processing on Compressed Data for Large Video Database. Proceedings of ACM Multimedia, pages 267--272, August, 1993.

....one such media instance we could integrate is the OVID system, though our framework is general enough to integrate many other media (which OVID cannot) The authors have developed feature identification schemes (which we have not) and this complements our work. In a similar vein, Arman et al. [2] develop techniques to create large video databases by processing incoming video data so as to identify features and set up access structures. Another piece of relevant related work is that of the QBIC (Query by Image Content) system of Barber et al. 3] at IBM, They develop indexing techniques ....

F. Arman, A. Hsu and M. Chiu. (1993) Image Processing on Compressed Data for Large Video Databases, First ACM Intl. Conf. on Multimedia, pps 267--272.

....similarity corresponds to minimum histogram difference. Similarity matrix S provides a quantitative measure that can be used to rank retrieval outcomes. Note that the similarity measure depends on the histogram differences that are already computed for scene change detection. Alternatively, in [5], clustering of camera shots based on the similarity measures of visual primitives such as color luminance correlation has been proposed. In [6] a time constrained clustering, which employs two similarity metrics to take into account both visual characteristics and temporal locality of shots, ....

F. Arman, A. Hsu, and M.Y. Chiu. Image processing on compressed data for large video databases. In Proc. 1st ACM Int. Conf. on Multimedia, pages 267-- 272, CA, 1993.

....3 Note that in the late 1800s, the concept of frames displayed at a pre specified rate was designed to support neither compression nor content based retrieval. It was introduced to force the human perception to observe motion. the physical representation of video to construct structure [FAC93] There have been efforts at the conceptual level with studies of appropriate data definition and manipulation languages for a video clip [OT93] based on the frame representation of a video clip. This dissertation is based on a top down approach. It proposes that a video clip maintains the ....

A. Hsu F. Arman and M-Y. Chiu. Image processing on compressed data for large video databases. In Proceedings of the ACM Multimedia, pages 267--272, 1993.

....[i; i 1) In contrast, by using segment trees instead, we allow leaves to have whatever granularity is needed to best represent the content of the video being annotated. Other works on video indexing include specific disk based techniques [3, 2] methods dealing with compression image processing [1], and synchronization structures [10] There has also been a good deal of work on picture retrieval systems [8, 7, 13] however, these works deal with static images, whereas in contrast we have studied ways of organizing temporal sequences of such images (i.e. video) 9 Conclusions Over the ....

F. Arman, A. Hsu and M. Chiu. (1993) Image Processing on Compressed Data for Large Video Databases, First ACM Intl. Conf. on Multimedia, pps 267--272.

....Most approaches are based on intensity histograms, and concentrate on cuts. Otsuji and Tonomura [11] discuss a variety of measures based on image differencing and changes in the image s intensity histogram. Nagasaka and Tanaka [10] present algorithms which use similar measures. Arman, Hsu and Chiu [2] have addressed the issue of change detection while operating directly on JPEG or MPEG encoded video; this approach is noteworthy for its efficiency. The above methods have difficulty with busy scenes, in which intensities change substantially from frame to frame. This change often results from ....

Farshid Arman, Arding Hsu, and Ming-Yee Chiu. Image processing on compressed data for large video databases. In Multimedia Conference, pages 267-- 272. ACM, 1993.

....a convolution algorithm for the DCT, and showed how it could be used for image processing in certain special cases. Their technique is applicable to high and low pass filtering, but does not adapt well to the block by block encoding nature of most compression technologies. 65 Lastly, Arman [4] has developed a technique for detecting scene breaks in motion JPEG video data that operates on the compressed representation. His technique compares two images using the dot product of two vectors formed from the DCT coefficients of a selected subset of corresponding block in the test images. If ....

F. Arman, A. Hsu, M. Y. Chiu, Image processing on compressed data for large video databases. Proceedings of ACM Multimedia 93, (Anaheim, CA, August 1-6, 1993). Association for Computing Machinery Press, New York, 1993 pp. 267-72.

....shot changes. Cherfaoui and Bertin [2] describe a technique based on camera movements. Zabih et al. 17] describe a technique using intensity edges that detects various edits such as cuts, dissolves, fades, etc. A lot of work has also been done on detecting shot changes in the compressed domain [1, 7, 9, 11, 13, 16, 19]. The earliest approaches were by Arman et al. 1] who used only the DCT information of I frames to detect disparity. This technique works well with clips containing only I frames (XING format or Motion JPEG) but has problems if the clip contains P and B frames and the I frames are spaced widely ....

....camera movements. Zabih et al. 17] describe a technique using intensity edges that detects various edits such as cuts, dissolves, fades, etc. A lot of work has also been done on detecting shot changes in the compressed domain [1, 7, 9, 11, 13, 16, 19] The earliest approaches were by Arman et al. [1], who used only the DCT information of I frames to detect disparity. This technique works well with clips containing only I frames (XING format or Motion JPEG) but has problems if the clip contains P and B frames and the I frames are spaced widely apart. Due to the reduction in temporal ....

[Article contains additional citation context not shown here]

F. Arman, A. Hsu, and M.Y. Chiu. Image processing on compressed data for large video databases. In Proc. of the ACM Multimedia Conference, pages 267-- 272, 1993.

....the relationships [30, 88] Information Extraction: Based on the data model ontology, concepts information are extracted and stored as metadata. Information can be extracted manually, automatically, or by combination of the two [7] Automatic extraction depends heavily on image processing tools [10, 13, 16, 37, 38, 39, 43, 49, 65, 67, 73, 94]. Information within unstructured data such as video is easily identified by human observation; however, few attributes can be identified by a machine. Therefore, we are more dependent on hybrid extraction techniques. User profile: A user profile represents information about user behavior and ....

F. Arman, A. Hsu, and M-.Y. Chiu, "Image Processing on Compressed Data for Large Video Databases," Proc. 1st ACM Intl. Conf. on Multimedia, Anaheim CA, August 1993, pp. 267-272.

....are developed and shot boundary extraction is carried out based on these models. These temporal video segmentation methods are independent of the type of the video and called syntactic methods. A number of syntactic temporal segmentation methods were also proposed in the compressed video domain [12], 13] 14] In addition to syntactic methods, domain dependent techniques have been proposed to describe the semantic characteristics of video streams; these methods rely on specific application data models to characterize the input data, and are aptly dubbed semantic methods [1] Swanberg et ....

F. Arman, A. Hsu, and M.Y. Chiu. Image processing on compressed data for large video databases. In Proc. 1st ACM Int. Conf. on Multimedia, pages 267--272, CA, 1993.

....of complex affine transforms, such as shearing and perspective mapping, into multipass operations. Content based search on images and video is another related area of research. Seales [14] has done work on object recognition using an eigenspace method in the compressed domain, and Arman[17] has shown how to perform cut detection in motion JPEG video data. 8. Conclusions In this paper, we have evaluated the performance of software implementations of compressed domain processing for the problem of MPEG to JPEG transcoding. We have also developed a novel practical method of ....

Farshid Arman, Arding Hsu, Ming-Yee Chiu, Image Processing on Compressed Data for Large Video Databases, Proceedings of the First ACM International Conference on Multimedia, Anaheim, CA, August 1993.

....which tend to ignore important low level features such as edges, corners or fine textures. There have been some preliminary edge extraction methods based on the classification of DCT coefficients. Some coarse edges can be extracted out directly from transform coefficients, e.g. Arman et al. [2] used the number of non zero coefficients in DCT block to decide whether there exists an edge. Shen and Sethi [17] used patterns of DCT coefficients to decide edge parameters such as orientation and strength. But neither of these can provide precise edge map To appear in Journal of Visual ....

F. Arman, et al, "Image Processing on Compressed Data for Large Video Databases," Proc. ACM Intl. Conf. Multimedia, June 1993.

....which consists of the start and finish point of maximal frame sequences fs1 and fs2. Output : An answer presentation fs1 N fs2 corresponding to the set of frames f(f 1 [ f n ) f 0 1 [ f 0 m )g int st, fi, start and end of each frame in the result sp[1] sp[2], variables pointing to array index dt[1] dt[2] variables for entering and exiting segment linenum; number of lines in consideration initialize linenum to zero ; initialize sp[1] and sp[2] to one ; initialize dt[1] and dt[2] to one ; set st to the min (ARR1[sp[1] ARR2[sp[2] ....

....of maximal frame sequences fs1 and fs2. Output : An answer presentation fs1 N fs2 corresponding to the set of frames f(f 1 [ f n ) f 0 1 [ f 0 m )g int st, fi, start and end of each frame in the result sp[1] sp[2] variables pointing to array index dt[1] dt[2], variables for entering and exiting segment linenum; number of lines in consideration initialize linenum to zero ; initialize sp[1] and sp[2] to one ; initialize dt[1] and dt[2] to one ; set st to the min (ARR1[sp[1] ARR2[sp[2] if (ARR1[sp[1] ARR2[sp[2] linenum = linenum ....

[Article contains additional citation context not shown here]

F. Arman, A. Hsu and M. Chiu. (1993) Image Processing on Compressed Data for Large Video Databases, First ACM Intl. Conf. on Multimedia, Anaheim, CA, Aug. 1993, pps 267--272.

....(EQN 21) EQN 22) EQN 23) The C code in figure 5 shows two functions, J and Jhat, that compute the operators defined in equations 19 and 20. The code uses two externally defined arrays, zz and qt, which encode the zig zag ordering and the quantization tables, respectively. zz[k] 0] and zz[k][1] specify the column and row, respectively, of the 8x8 block element corresponding to the element in the zigzag vector indexed by k, and qt[u] v] is the standard JPEG quantization table. 3 LGDVEs on JPEG Data Having formulated LGDVEs, and significant parts of JPEG compression and decompression as ....

....domain operator. Their method approximates the coefficients in the compressed domain operator using powers of two, allowing the result to be computed using only shifts and adds. Lastly, Arman developed a technique to detect scene changes in motionJPEG compressed video data in the DCT domain [1]. Many extensions are possible to the work presented in this paper. Condensation algorithms can be developed that improve the overall image quality using better metrics for finding insignificant coefficients than thresholding condensation. For example, the first author s dissertation [23] explored ....

F. Arman, A. Hsu, M. Y. Chiu, Image Processing on Compressed Data for Large Video Databases, Proceedings of the First ACM International Conference on Multimedia, August 1993, Anaheim, Calif.

....to refer to a sequence of video representing continuous action and story to refer to a sequence of scenes. The problem of identifying cuts (sharp transitions between shots) has been typically approached from a bottom up perspective, looking for rapid changes in color histogram or image intensity (Arman, 1993; Nagasaka, 1992; Zhang, 1993) Model based algorithms have also been developed to successfully detect fades, dissolves, and page translate edits (Hampapur, 1994) Once shots have been identified, keys frames which characterize the shot can be selected by considering the motion of objects within ....

Arman, F., Hsu, A., et al, (1993). Image Processing on Compressed Data for Large Video Databases, ACM Multimedia '93, California, USA, 267-272.

....(DCT) coefficients of each MB, and the motion vector components for the forward, backward, and bidirectionally predicted MBs, we reduce the need for decompression. Detecting some physical changes such as cuts and camera motion is fairly easy and algorithms have appeared in many recent papers [2, 8, 16, 19]. Detecting gradual transitions and special effect edits, on the other hand, is a tougher problem in the compressed domain. We have developed a structural representation of a video clip that can be used to tackle such problems. 1.2 Related Work As mentioned earlier, a great deal of work has been ....

F. Arman, A. Hsu, and M.Y. Chiu. Image processing on compressed data for large video databases. In Proc. of the ACM Multimedia Conference, pages 267--272, 1993.

....has also been a great deal of work on using computers to automatically extract video content. The simplest and most common technique is video segmentation (i.e. detecting the boundaries between camera shots) Recent research in this area includes work on shot detection using compressed video [1] and work on automatic classification of shot transitions [31] Swanberg has used a knowledge base to do automatic segmenting and scene classification for a highly structured type of video [28] Research has also been done on object tracking and classification of object motion in video [11] While ....

Arman, F., Hsu, A., and Chiu, M-Y., "Image Processing on Compressed Data for Large Video Databases", Proc. ACM Multimedia 93, Anaheim, CA, August, 1993.

....which contained an anchor person, outdoor and indoor scenes. Various transitions, such as simple cuts, wipes, fades, and animated graphics, were included as well. Figure 3 exhibits the results of calculating 1 y for a few minutes of a JPEG encoded video of a news type broadcast (see also [2]) In Figure 3 a, the time step, j, is set to 64. With relatively few calculations, several of the time periods in which no transitions occur are identified most notably from frame number 1250 to 1750. As mentioned earlier, j is divided by two when 1 y t 1 (set to 0.55) and y is ....

F. Arman, A. Hsu, and M-Y. Chiu. Image processing on compressed data for large video databases. In Proceedings of First ACM International Conference on Multimedia, pages 267--272, Anaheim, CA, 1-6 Aug. 1993.

No context found.

F. Arman, A. Hsu and M.Y. Chiu, "Image processing on compressed data for large video databases", Proc. ACM Multimedia 93, pp. 267-272, Anaheim, CA, 1993.

No context found.

F.Arman, A. Hsu, M.Y. Chiu: "Image processing on compressed data for large video databases", ACM Multimedia 1993, pages 267--272

No context found.

F. Arman, et al, "Image Processing on Compressed Data for Large Video Databases", Proc. ACM Intl. Conf. Multimedia, pp. 267-272, June 1993.

No context found.

F. Arman, A. Hsu, and M.Y. Chiu. "Image processing on compressed data for large video databases". Proc. of the ACM Multimedia Conference, pages 267--272, 1993.

No context found.

F. Arman, A. Hsu, and M.Y. Chiu. Image processing on compressed data for large video databases. Proc. of the ACM Multimedia Conference, pages 267--272, 1993.

No context found.

F. Arman, A. Hsu, and M. Chiu, "Image processing on compressed data for large video databases," in Proc. of the ACM Multimedia Conference, pp. 267--272, 1993.

No context found.

F. Arman, A. Hsu, and M. Chiu, "Image processing on compressed data for large video databases," in Proc. of the ACM Multimedia Conference, pp. 267--272, 1993.

No context found.

F. Arman, A. Hsu, M. Chiu: "Image processing on compressed data for large video databases". Proc. of the ACM Multimedia, pp. 267-272, California, USA, June 1993.

No context found.

F. Arman, A. Hsu, , and M.-Y. Chiu, "Image processing on compressed data for large video databases," in Proc. First ACM Int'l Conf. Multimedia, (Anaheim, CA), August 1993.

No context found.

F. Arman, A. Hsu and M.-Y. Chiu, "Image processing on compressed data for large video databases," Proc. of ACM International Conference on Multimedia, pp. 267-272, Anaheim, CA, USA, 1993.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC