Title of Invention

METHOD AND DEVICE FOR ENCODING OF MULTIMEDIA OBJECT

Abstract

Coding is provided in which a multi-media object is coded to obtain a bit-stream, and quality information is added to the bit-stream, which quality information indicates a quality of the object in relation to a given position in (or a given part of) the bit-stream. By adding quality information to the bit-stream, jointly storing or transmitting multiple coded objects can be optimized in that the quality of the object can be easily taken into account. (fig 1)

Full Text

The invention relates to coding multi-media objects-Scalable compression, e.g. fine-granularity scalable compression of objects such as multi-media objects has the useful feature that the encoded bit stream may be truncated at a given point, while the remaining stream can still be decoded (although at a lower object quality). A standard of such a scalable coding, i.e. MPEG-4 Fine Granularity Scalability, is currently being defmed, see ISO/EC 14496-2 / AMD 4, document ISO/EC JTC1/SC29AVG11 N3315, March 2000 (further called N3315), which is incorporated by reference herein. A further scalable coding method is described in non pre-published European Patent Application 00201037.9, filed 2000.03.23 (our reference PHNL000153), which is also mcorporated by reference herein. The availability of such a scalable bit stream considerably simplifies system designs by practically eliminating the need for a buffer control method when fitting the encoded bit stream to a certain given bit rate or memory size. In particular, the same single bit stream simultaneously serves different channels with different capacities, without the need to re-encode the original data. Thus, real-time adaptation to varying channel capacities (with application to the Intemet or wireless communication chaimels) is very much simplified. Before fine granularity scalability, already some forms of limited scalability existed. There, the bit stream consisted of a few large layers, i.e. a base layer and e.g. one or two enhancement layers. Such scalability is defined e.g. in the JPEG standard (hierarchical coding) as well as in the MPEG2 standard (SNR scalability, spatial scalability, temporal scalability). An object of the invention is to provide advantageous coding. To this end, the invention provides coding of a multi-media object to obtain a bit-stream, controlling a bit-stream, transmitting a bit-stream, receiving a bit-stream, a multiplexer or network node, a (scalable) bit-stream representing a multi-media object, a storage medium, a computer program, and a signal carrying a computer program. According to a first aspect of the invention, a multi-media object is coded to obtain a bit-stream, and quality information is added to the bit-stream, which quality information indicates a quality of the object in relation to a given position in (or a given part of) the bit-stream, wherein quality tags are added to the bit-stream, a quality tag providing a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a given point related to the quality tag. By adding quality information to the bit-stream, jomtly storing or transmitting multiple coded objects can be optimized in that the quality of the object can be easily taken into account. This aspect of the invention is based on the insight that it is easy to determine the rate of a compressed object, but that another important parameter, a quality measure, e.g. distortion, is not so easily determined. In fact, the distortion can only be accurately obtained at the time of coding, when the complete source information is still available. According to this aspect of the invention, the bit stream syntax is enhanced by adding quality (distortion) information by providing quality tags having a quality tag value which quality tag value represents a quality of the reproduction of the encoded object when the bit-stream is truncated at a given point related to the quality tag. This can be done at no or a negligible increase in bit rate and extends the range of applications for several coding schemes. The multi-media object may be an audio and/or video object or any other reproducible object for which a quality is relevant. The multi-media object may also be a picture or a sequence of pictures such as a program. Preferably, the coding is a scalable coding and the resulting bit-stream is a scalable bit-stream. Especially for scalable coding schemes, quality information is advantageous because these bit-streams are suitable for truncating. For a lot of applications, wherein scalable bit-streams are truncated, it is important to have a quality indication of the bit-stream resulting afler truncation, which is easily provided by the quality information comprised in the quality tag value included in the scalable bit-stream. The quality tag value represents object reproduction quality when the object is truncated at a point related to the quality tag. Information on object reproduction quality versus number of bits, i.e. the truncation point, is then easily determined. To quantify the quality, preferably signal-to-noise ratio (SNR) or peak-signal-to-noise-ratio (PSNR) values are used Whereas the encoded stream of a single object may be truncated optimally by just fitting it to the available bandwidth/storage, this is not the case when simultaneously dealing with multiple objects. To optimally allocate a certain bandwidth or storage space to multiple objects simultaneously, one has to know the differential rate-distortion curve for each encoded object. While this curve is relatively easily obtained during the encoding of an object (when the original is available), it is non-trivial to obtain (by estimation) later, when only a truncated version of the encoded bit stream is available. Estimation of the quality at a later time requires detailed knowledge of the compression method as well as at least partial decoding of the encoded bit stream. The quality tag values of the quality tags added to the scalable bit-stream represent the quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to a given tag. Although the addition of the quality information may require a given overhead, this overhead can be kept small An important advantage is that the quality information makes it easy to jointly optimally truncate the bit-streams of multiple objects. Such a multiple truncation problem occurs for example in an elastic memory as described in non pre-published European Patent Application 00200890.2, filed 2000.03.13 (our reference PHNLOOOllO), which is incorporated by reference herein. Another application is a multiplexer or a network node in which the outgoing bandwidth is temporarily lower than the incoming bandwidth and consequently the incoming scalably compressed bit-streams need to be truncated. When doing compression, the input data is usually compressed in multiple units (such as e.g. parts of DCT blocks, parts of frequency bands of a wavelet transformed image or layers). Each coded part usually contains some headers with various parameters or tags. In an advantageous embodiment of the invention, in such a header, a parameter is added indicating the quality of the object when it is truncated just after (or altematively just before) the current encoded data part. One example of a quality parameter is to add a number related to the mean squared error (or PSNR or SNR) of the reproduction; the number might also represent a visually weighted (P)SNR. The type (or multiple types) of quality indication might be standardized, so all encoders will use the same or a limited number of different quality indicators. The quality could also be relative (for example a percentage), so an encoder would not have to disclose its quality measure. The relative quality might then range from 0 to 100% of a certain scale for each individual object, with an additional scale/weight parameter for each object to enable different weighting of the various objects. The quality tags may be placed at approximately equal distances (number of bits) throughout the encoded stream or they may be used more frequently when the bit rate versus quality curve is quickly changing. When an encoded bit stream has to be truncated, the quality for each truncation point not corresponding to a tag location can be approximated by interpolation (linear or more complexaO of the quality tag values. In an embodiment of the invention, the quality information is added to the encoded bit stream of MPEG-4FGS using a tag that has already been defined in the standard, see the above-mentioned reference N3315. In this way, the quality information can be added without having to change the proposed standard, which is a tremendous advantage. For decoded multi-media objects, the quality information may be used for adaptive post-processing or for scalable video processing algorithms etc for non-scalable compression methods. For example, for post-processing of MPEG-compressed video, the quality information can help to determine the 'strength' or amount or the (blocking artifacts/ ringing reduction) of post-processing required. For scalable video aigorithms, the quality information can help to better estimate the number of CPU cycles required to achieve a certain desired processing quality level using a certain selected video processing algorithm. The quality information may be added as side information to the bit-stream, i.e. not included in the bit-stream itself For encrypted bit-streams, it is advantageous that the quality information is unencrypted. The quality of a given part of the bit-stream (e.g. layer) can then be determined in a decoder without decrypting the bit-stream. Quality information can also be advantageously applied for applications in which source coding and channel coding are not carried out at the same time or location. The quality information is then used in the channel coding, e.g. to determine the protection rates. The aforementioned and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. In the drawings: Fig. 1 shows a system according to an embodiment of the invention, and Fig. 2 shows more advantageous embodiments of the invention. The drawings only show those elements that are necessary to understand the invention. Fig. 1 shows a system according to an embodiment of the invention, the system comprising a transmitter 11 having an input umt or object generation unit 110 and an encoder 12. The encoder 12 comprises a scalable encoder 120 and a quality information generation unit 121. The scalable encoder 120 codes objects obtained from the input unit 110 to provide one or more scalable bit-streams. The quality information generation unit 121 extracts the object quality from the signals obtained from the input unit 110 as well as the signals and/or parameters provided by the encoder 120. The quality information from the generation unit 121 is provided to the encoder unit 120, which generates the quality information tags and inserts them in the scalable bit-stream. The system further comprises a truncator 3 for truncating the one or more scalable bit-streams, and a truncator control vmit 4. The truncator control unit 4 extracts quality information from the scalable bit-stream provided by the encoder 12 and controls the truncator 3 in dependence on the received quality information/tags. In the case of only one scalable bit-streaia, the scalable bit-stream is truncated when the desired quality has been reached. Truncator 3 and comntrol unit 4 togedier may constitute part of a multiplexer, bit-rate control unit, network node, etc. and may be present in a channel, but also in a receiver. Unit 5 may alternatively be a reproduction unit and/or decoder, e.g. being present together with truncator 3 and control unit 4 in a receiver according to an embodiment of the invention. In more advantageous applications, as shown in Fig. 2, multiple scalable bit-streams are provided by transmitters 21,31,41, wherein at least some of the multiple scalable bit-streams have quality tags included in them. The transmitters 21531,41 and their components are similar to transmitter 11 shown in FIG. 1. Depending on the available bandwidth or storage capacity on a channel or storage medium 15, the scalable bit-streams are more or less truncated, under dependence of the quality information/tags that are present in the scalable bit-streams. Such a multiple truncation can be done using the principle of elastic memory described in non pre-published European Patent Application 00200890.2, filed 2000.03.13 (our reference PHNLOOOl 10), which is incorporated by reference herein. Multiplexer 16 combines the streams from the transmitters. Truncator 13 and control unit 14 together may constitute part of a multiplexer (e.g. 16), bit-rate control unit, network node, etc. and may be present in a channel, but also in a receiver. Unit 15 may alternatively be a reproduction unit and/or decoder, e.g. being present together with truncator 13 and control unit 14 in a receiver according to an embodiment of the invention. In the following, some examples of applications of MPEG-4 FGS that need the quality infoin:iation are given. Although the following is addressed in particular to MPEG-4 FGS, it will be clear to a person skilled in the art that the invention can be advantageously apphed to any scalable coding scheme. From an application point of view, the distortion is a significant parameter for the MPEG-4 FGS scheme. If distortion information is not available, the usability of FGS is limited, as is demonstrated below by giving various applications that do need this information. According to an embodiment of the invention, the FGS bit stream syntax is enhanced by adding quality (distortion) information. This can be done at no or a negligible increase in bit rate and extends the range of applications for FGS. A first application of the invention is the coding for a constant-quality (thus variable bit rate) output. This can be used, for example, for recording video data with constant quality on a storage medium that allows for a variable bit rate. Using the quality information, the final bit stream does not need to be produced during the initial encoding but it can be obtained by processing the encoded bit stream at a later time. Selling the same content at different qualities can be efficiently implemented using a scalable (fme granularity or layered) compression method such as described above followed by encryption of one or more of the layers: a property of many scalable compression methods is that when the lowest scalability layer is not available, the higher scalability layers are useless, i.e. cannot be used to increase the quality. When the scalably compressed content is encrypted, it is still possible to use it for elastic storage, i,e. to reduce the amount of storage space by throwing away some of the enhancement Iayer(s). For elastic storage, reference is made to PHNLOOOllO as mentioned before. To decide how much data to remove, some information about the associated quality loss should be available, since this information can no longer be derived from the compressed bit stream without decrypting it. In the current embodiment of the invention, the quality information is sent as unencrypted information, e.g. as side information. The business model of selling tihe same content at different quality levels is closely related to elastic storage, since there too the same content is stored at multiple quality levels using scalable compression. The quality levels that are offered for purchase to the consumer preferably directly correspond to the quality levels used in the elastic storage system. This implies that when the elastic storage device wants to lower the quaUty of a certain content item, it can remove the highest encrypted quality layer, without needing to decrypt it. Since the device does thus not decrypt any data, there is no security or theft risk. To maintain security in the whole chain from content owner or service provider to consumer, the content is preferably compressed (using a scalable compression method) and encrypted at the desired quality levels by the content owner and then distributed in encrypted form to the elastic storage device (either directly transmitted or downloaded or indirectly via e.g. intemiediate storage on an optical disk). In an elastic storage application, the user (or the device based on what it knows about the preferences of the user) may optionally select a certain desired minimum quality level. I.e. content that is currently available at a higher quality level than minimally desired by the user may be reduced in quality, to make room for more different content, until it reaches the lowest quality acceptable to the user. When the content is still available in a higher quality, however, the user still has the option to purclmse the higher quality. Of course the user (or device) may also preset different desired minimum quality levels for different types of content (like sports, talk shows, or movies). Alternatively, it is also possible to let the service provider manage the storage space and determine which qixality levels should be removed (in that case, the service provider keeps track of the qualities and may carry out the elastic storage functions). This could be useful e.g. when content is put on a set-top box containing a storage function (e.g. hard disk) by the service provider. Initially, the content could be offered to the user at a high quality. When the user does not watch/buy the content within a certain time, the quality level stored on the set-top box could be lowered to make room for different content. The quality information also allows source encoding and channel coding to be carried out at a different time or location. This is useful because at the time of aicoding the channel characteristics may not yet be known. Also, the same encoded bit stream may serve different channels v\ath different error characteristics. Finally, no storage space is wasted for storing the error correction overhead. It can be generated when needed, since the quality fields give the required information for adding the channel coding (using unequal error protection). Then there are applications where multiple encoded FGS frames have to be jointly processed. This can occur, for example, in a congested network node, wheretemporarily less bandwidth is available. The network node can then use the quality information to optimally truncate the bit streams with the minimal loss of quality. Because multiple objects are involved, with possibly very different rate-distortion curves, the truncation cannot be satisfactorily done without the quality information. Additionally, for streaming applications the quality information can provide the server with a good tool to perform the rate-control at transmission time and also the tradeoff between SNR and temporal enhancements (FGS versus FGST, see for definitions N3315). In MPEG-4 FGS, the quality information that is needed is actually the rate-distortion curve for the scalable enhancement layer. Since the rate is obvious, only the distortion information has to be added. Twosolutions are proposed that allow adding this information with no or minimal modifications of the current FGS bit stream syntax. Solution A. The start of a bit plane is a good point for adding quality information/fields, because it allows to easily retrieve the information and also provides sufficient samples to accurately describe the rate-distortion curve. The "quality code" would be similar to the current fgs_bp_start__code, whose last 5 bits indicate the ID of the bit plane. Instead of the bit plane ID, the quality information can be inserted in these 5 bits. In a ferred embodiment, a bit is added to the syntax to indicate whether the fgs_bp_start_code itains either the bit plane ID or the quality information. Solution B. Alternatively, a new code could be defined for the quality ormation and be inserted after the fgs_bp__start_code. In this case, the quality tag can we 8 bits. The information we propose to store in the quality field is the distortion after npletely decoding the bit plane following the quality field In this way, when the stream is ncated inside a bit plane, the approximate quality may be obtained by interpolation. This is lier than the extrapolation that would be required if the quality field were to contain ormation about the distortion before decoding the current bit plane. For example, let Ql be quality before decoding a bit plane and Q2 the quality after decoding it (as proposed we, Q2 is known to the decoder aheady at the start of the bit plane). If the stream is ncated inside the bit plane, it is thus known that the true quality Q lies inside the interval |1,Q2>. It can therefore simply be approximated as Q ^ (Ql-rQ2);, A more accurate Koximation can be made by also taking into account the number of decoded EKJT blocks. r example, if the enhancement information for the current bit plane has been received for n L of the N total blocks for a frame, the true quality can be approximated as Q = Q1+(Q2 -)*n/N. In a preferred embodiment for Solution A, a first quality field for an lancement VOP, i.e. the field for the most significant (MSB) bit plane, contains an jolute quality (distortion), whereas the additional fields contain quality improvements stortion reductions) relative to the previous quality. The absolute quality can be used to compare different objects. Putting quality improvements in the additional fields allows these improvements to be represented with a higher accuracy then when absolute qualities would be used. This is particularly important when only 5 bits are available for each field. To quantify the quality, preferably PSNR values are used. As stated above, the first quality field contains the absolute quality. We propose to use the 5 bits to give the PSNR after decoding the first (MSB) bit plane, witii a range of 18...49 dB in steps of 1 dB. This range covers all practically relevant PSNR values: when the PSNR is above 49 dB, the base layer already contains a near-lossless representation of the object. A PSNR that is lower than 18 dB would mean the base layer provides an extremely low quality, which is not very likely. When values outside the range do occur, they will be clipped to either 18 or 49 dB, depending on whether they fall below or above the allowed range. The next quality fields will then contain the improvement in quality for completely decoding the following bit plane, relative to the quality after decoding the previous bit plane. Preferably, the 5 bits are used for giving these quality improvements the range of 0.. ,6.2 dB in steps of 0.2 dB. Since a single bit is added, the improvement cannot exceed 6.02 dB, so this range is sufficient. For Solution B, when 8 bits are used for the quality tag, the quality values (both absolute and differential) could be represented with even finer grain. However, Solution B would also allow us to simply use only absolute (i.e, non-differential) quality values. The quality range would then be 18.. .60.5 dB in steps of 1/6«0.167 dB (or 18..,49.875 dB msteps of 0.125 dB). Various appUcations of MPEG-4 FGS have been discussed above, which application need quality information. Since this information is only completely available when the original encodiag takes place, it is added to the bit stream to make it available for later use. This can be done at no or a negligible increase in bit rate with minimal modifications of the current bit stream syntax. Two detailed solutions have been presented for adding the PSNR quality values. Solution B using absolute quality values is preferred. The invention applies to all cases where multiple scalably compressed multimedia objects have to be jointly stored or transmitted and some of these objects have been compressed by MPEG-4 FGS incorporating the invention. Particular applications are the elastic memory applications as well as transmission channels or networks dealing with multiple objects/users. When the memory/chaimel/network has to be shared by few objects/users they get a high quality. The quality is automatically reduced to accommodate more objects/users. This can be done efficiently, i.e. with low overhead, because of the presence of the quality tags according embodiments of this invention. The invention can also be advantageously applied for applications in which source coding and channel coding are not carried out at the same time or location. The quality tags then give the required information for adding the channel coding (equal error protection, e.g. more protection for parts of the bit-stream that represent higher quality, or more protection for parts of the bit-stream with a high quality to number of bits ratio). The invention may also be advantageously applied in the context of scalable image processing schemes such as JPEG2000, see document ISO/IEC JTC 1/SC 29AVG 1 N1646, dated 16 March 2000, which is iucorporated by reference herein. The quality infonnation may conveniently be included in JPEG2000, because a Comment and Extension Marker (CME) has already been defined (see page 51 of document N1646), which allows unstructured data in the header. Quality information is advantageously included in a given CME. E.g. binary data can be included (Rcme=0). Further, according to an CTobodiment of the invention, a separate Rcme type is defined for quality tags. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word 'comprising' does not exclude the presence of other elements or steps than those listed in a clabn. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a device claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. 1. A method of encoding a multi-media object, the method comprising the steps of: coding the object to obtain a bit-stream, and adding quality information to the bit-stream, which quality information indicates a quality of the encoded object in relation to a given position in (or a given part of) the bit-stream, wherein quality tags are added to the bit-stream, a quality tag providing a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a given point related to the quality tag. 2. The method as claimed in claim 1, wherein the coding step is a scalable coding step to obtain a scalable bit-stream. 3. The method as claimed in claim 1, wherein the quality tag value is based on a signal to noise ratio value. 4. A method as claimed in any of the preceding claims, wherein quality tags are added at given locations in the bit-stream, the quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated just after (or altematively just before) the given location in the bit-stream. 5. The method as claimed in claim 1, wherein the quality tag is incorporated in existing fields of a given scalable coding standard. 6. The method as claimed in claim2, wherein the scalable bit-stream includes several layers and wherein respective layers include respective quality tags. 7. The method as claimed in claim 1, wherein the bit-stream is encrypted and the quality information is unencrypted. 8. A method of controlling at least one bit-stream representing a multi-media object encoded by a method as claimed with any of the preceding claims 1 to 7 in which bit-stream quality information has been added in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the quality tag, the method comprising the steps of: receiving the at least one bit-stream, extracting the quality tag value from the quality tags for the bit-stream, transcoding or truncating the at least one bit-stream in the case a desired combination of bit-rate and reproduction quality of the at least one bit-stream differs from a current combination of bit-rate and reproduction quality of the at least one received bit-stream, providing the at least one transcoded or truncated bit-stream at the desired combination of bit-rate and reproduction quality. 9. A method of transmitting at least one multi-media object, comprising transmitting a encoded bit-stream encoded by means of a method as claimed in any of the claims 1 to 7 in which the quality information has been added in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the given quality Jag. 10. A method of receiving at least one bit-stream representing a multi-media object in which bit-stream quality information has been added, by means of a method as claimed in any of the claims 1 to 7, in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the given quality tag, the method comprising the steps of: extracting the quality information from the quality tag values in the bit-stream, transcoding or truncating the at least one bit-stream in the case a desired combination of bit-rate and reproduction quality of the at least one bit-stream differs from a current combination of bit-rate and reproduction quality of the at least one received bit-stream, providing the at least one transcoded or truncated bit-stream at the desired combination of bit-rate and reproduction quality, and decoding the at least one bit-stream at the desired combination of bit-rate and reproduction quality. 11. A method of receiving at least one bit-stream representing a multi-media object in which bit-stream quality information has been added, by means of a method as claimed in any of the claims 1 to 7, in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the given quality tag, the method comprising the steps of: extracting the quality information from the quality tag values in the bit-stream; decoding the bit-stream to obtain a decoded multi-media object; and processing the multi-media object in dependence on the extracted quality tag values. 12. A device of encoding a multi-media object, the device comprising: means for coding the object to obtain a bit-stream, and means for determining for one or more points in the coded bit-stream a quality value representing the reproduction quality of the object when the encoded object is truncated at said one or more points, means for adding quality information to the bit-stream in the form of quality tags related to said one or more points, said quality tags having a quality tag value representing the reproduction quality of the object when the encoded object is truncated at the point related to the quality tag. 13. A transmitter comprising a device as claimed in claim 12. 14. A controller for controlling at least one bit-stream representing a multi-media object in which bit-stream quality information has been added, by means of a method as claimed in any of the claims 1 to 7, in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the given quality tag, the controller comprising: means for receiving the at least one bit-stream, means for extracting the quality information from quality tag values in the bit-stream, means for truncating the at least one bit-stream in the case a desired combination of bit-rate and reproduction quality of the at least one bit-stream differs from a current combination of bit-rate and reproduction quality of the at least one received bit-stream, means for providing the at least one truncated bit-stream at the desired combination of bit-rate and reproduction quality. 15. A receiver comprising a controller as claimed in claim 14. 16. A receiver for receiving at least one bit-stream representing a multi-media object in which bit-stream quality information has been added, by means of a method as claimed in any of the claims 1 to 7, in the form of quality tags, a quality tag having a quality tag value representing a quality of the reproduction of the encoded object when the bit-stream is truncated at a point related to the given quality tag, the receiver comprising: means for extracting the quality information from quality tag values in the bit-stream; means for decoding the bit-stream to obtain a decoded multi-media object; and means for processing the multi-media object in dependence on the extracted quality tag values.

Documents:

abs-in-pct-2002-874-che.jpg

in-pct-2002-0874-che abstract duplicate.pdf

in-pct-2002-0874-che claims duplicate.pdf

in-pct-2002-0874-che description (complete) duplicate.pdf

in-pct-2002-874-che- abstract.pdf

in-pct-2002-874-che- claims.pdf

in-pct-2002-874-che- correspondence others.pdf

in-pct-2002-874-che- correspondence po.pdf

in-pct-2002-874-che- descripition complete.pdf

in-pct-2002-874-che- drawings.pdf

in-pct-2002-874-che- form 1.pdf

in-pct-2002-874-che- form 26.pdf

in-pct-2002-874-che- form 3.pdf

in-pct-2002-874-che- form 5.pdf

in-pct-2002-874-che- other documents.pdf

in-pct-2002-874-che- pct.pdf