T 0762/12 () of 8.3.2013

Summary of Facts and Submissions

I. The appellant appealed against the decision by the examining division refusing European patent application No. 04 100 907.7.

II. The present application claims priority from three applications, the earliest being US patent application:

P1: US 60/026,963 P filed on 20 September 1996.

III. The examining division refused the application on the ground that the subject-matter of the claims in the requests then on file was not entitled to the earliest priority from P1, so that the subject-matter of the claims was not novel over document:

D1: PURI A. et al.: "Improvements in DCT Based Video Coding", PROCEEDINGS OF THE INTERNATIONAL SOCIETY FOR OPTICAL ENGINEERING, SPIE, Vol. 3024, 12 February 1997, pages 676 to 688, XP000199866, ISSN: 0277-786X.

The examining division further commented as an obiter dictum that, should the two earliest priority claims not be valid for the present application, the subject-matter of the claims might also not be novel in view of document EP 0 843 484 A1.

IV. The following further documents were cited in the proceedings:

D2: PLOYSONGSANG A. et al.: "DCT/DPCM Processing of NTSC Composite Video Signal", in IEEE TRANSACTIONS ON COMMUNICATIONS, USA, Vol. COM-30, No.3, March 1982, pages 541 to 549, XP002062314, ISSN 0090-6778;

D3: US 2,905,756 A;

D6: MUSMANN H.G. et al.: "Advances in Picture Coding", in PROCEEDINGS OF THE IEEE, Vol. 73, No. 4, April 1985, pages 523 to 548, XP002057834, ISSN: 0018-9219;

D7: TAN T.K. et al.: "Intra Prediction (T9/T10) and DC/AC Prediction Results", ISO/IEC JTC1/SC29/WG11, No. MPEG96/0939, 30 June 1996; XP030030333, ISSN 0000-0331;

D8: GRAHAM R.E.: "PREDICTIVE QUANTIZING OF TELEVISION SIGNALS", IRE Wescon, Convention Record, Part 4, Vol. 2, No. 4, 19 August 1958, pages 147 to 157, XP002051791;

D9: NETRAVALI A.N. et al.: "Picture Coding: A Review", PROCEEDINGS OF THE IEEE, Vol. 68, No. 3, March 1980, pages 366 to 407, XP002028499, ISSN: 0018-9219;

D10: COHEN P. et al.: "ADAPTIVE DIFFERENTIAL CODING OF PICTURE SIGNALS BASED ON LOCAL CONTOUR PREDICTION", COMMUNICATIONS AND KNOWLEDGE, PARTNERS IN PROGRESS, 29 November to 1 December 1976, Vol. 1, pages 6.1-1 to 6.1-5, XP002039931;

D11: GONZALEZ R.C. et al.: "Digital Image Processing", third edition, Prentice Hall, 2008, pages 543, 566 to 573, 580 to 584, 596 to 602 and 621 to 622;

D13: PENNEBAKER W.B.: "JPEG Still Image Data Compression Standard", New-York, 1993; two cover pages and pages 29 to 37; 11 pages in total;

D14: printout dated 07—10—2010 of

http://mpeg.chiariglione.org/about_mpeg.htm, two pages;

D15: ISO/IEC JTC1/SC 29 N 1963, "Coding of Audio, Picture, Multimedia and Hypermedia Information", 12 March 1997, seven pages.

V. In an annex to the summons to oral proceedings, the board raised the further issue of lack of inventive step, starting from either document D2 or document D7.

VI. Oral proceedings were held jointly on 7 and 8 March 2013 in appeal cases T 0762/12 and T 0763/12, both relating to divisional applications from the same parent European patent application No. 97 307 378.6.

VII. The appellant requested that the decision under appeal be set aside and that a patent be granted on the basis of claims 1 to 7 according to the sole request filed in the oral proceedings on 8 March 2013.

VIII. Independent claim 1 reads as follows:

"A decoding method for decoding MPEG-4 video signals associated with a block X of 64 coefficients, the method comprising:

receiving MPEG-4 video signals comprising a plurality of blocks of data, each block comprising a quantised DC residual signal,

determining a first gradient between a quantised DC coefficient DCA of a previously decoded block A that is horizontally adjacent to and left of the block X, and a quantised DC coefficient DCB of a previously decoded block B that is vertically adjacent to and above the block A;

determining a second gradient between DCB and a quantised DC coefficient DCC of a previously decoded block C that is vertically adjacent to and above the block X, wherein each of the block A, the block B and the block C is inside a boundary of a video object plane containing the block X;

when the first gradient is less than the second gradient, predicting a DC coefficient DCX of the block X from DCC, otherwise, predicting the DC coefficient DCX from DCA;

identifying the residual signal for the block X from the MPEG-4 video signals;

generating a DC coefficient for the block X from the predicted DC coefficient DCX of the block X and the residual signal for the block X; and

generating image data for the block X according to an inverse discrete cosine transform based on the generated DC coefficient."

IX. Independent claim 5 reads as follows:

"A video decoder for decoding an MPEG-4 formatted video bitstream associated with a block X of 64 coefficients, the decoder being arranged:

to receive MPEG-4 video signals comprising a plurality of blocks of data, each block comprising a quantised DC residual signal,

if |DCA-DCB|<|DCB -DCC|, to predict at a DC coefficient predictor (400) a DC coefficient of the block X from DCC, otherwise, to predict a DC coefficient DCX of the block X from DCA, where DCA is a quantised DC coefficient for a previously decoded block A, DCB is a quantised DC coefficient for a previously decoded block B, DCC is a quantised DC coefficient for a previously decoded block C, the block A is horizontally adjacent to and left of the block X, the block B is vertically adjacent to and above the block A, and the block C is vertically adjacent to and above the block X, wherein each of the block A, the block B and the block C is inside a boundary of a video object plane containing the block X,

to identify the residual signal for the block X from the MPEG-4 video signals;

to generate at an adder a DC coefficient for the block block [sic] X from DCX and the residual signal for the block X; and

to generate image data for the block X according to an inverse discrete cosine transform based on the DC coefficient."

X. The reasons in the decision under appeal with respect to the entitlement to the earliest priority from P1 and regarding lack of novelty may be summarised as follows:

P1 describes a very specific method using predicted DC and AC transform coefficients to determine gradients and to predict coefficient DCX. P1 does not contain claims. Generalising the specific method disclosed in P1 as applicable to any (unquantised) transform coefficients is not directly and unambiguously derivable from P1. Furthermore, predicting coefficient DCX "based on" or "using" either coefficient DCC or coefficient DCA is also an unsupported generalisation of the specific formulas disclosed in P1. Hence the independent claims of the requests then on file are not entitled to the priority from P1 (see also decision G 2/98), so that document D1 belongs to the state of the art within the meaning of Article 54(1) and (2) EPC and deprives their subject-matter of novelty.

XI. The objection of lack of inventive step raised by the board in the communication accompanying the summons and in the oral proceedings may be summarised as follows:

The decoding method of claim 1 differs from the method known from the closest prior-art document D2 by the neighbouring blocks of data used for predicting the DC coefficient of block X.

This difference solves the problem of implementing an alternative block-based predictive decoding method. The solution according to claim 1 is known in the prior art as Graham's rule (see documents D3, D6, D8 or D9). The skilled person would immediately recognise that the rule is applicable to pixel prediction in the spatial domain as well as to prediction of the DC coefficient in the frequency domain.

Adopting this solution in the method of D2 is obvious. The same would apply starting from document D7.

XII. The arguments by the appellant may be summarised as follows:

The amended claims are entitled to priority from P1. Thus D1 does not belong to the state of the art.

The decoding method of claim 1 differs from the decoding method of the prior art D2 by:

- decoding MPEG-4 signals;

- predicting the DC coefficient based on the result of the comparison of differences between previously decoded DC coefficients.

The technical problem should be formulated generally as providing an alternative block-based coding scheme.

D2 relies on decorrelation (see also D11 and D13). The DC coefficient of one subblock only represents the average intensity of that subblock and thus does not represent the average intensity of spatially adjacent pixels, whereas D3 relies on spatial correlation between adjacent pixels. D2 and D3 are thus fundamentally incompatible.

DC coefficient prediction is based in D2 on a subblock to be predicted and four neighbouring subblocks, and requires the transmission of overhead information. The skilled person would not have considered selecting a less accurate method, such as Graham's rule, which uses only values of previously received adjacent pixels and does not require overhead information transmission.

In conclusion, there was no incentive at the filing date of the present application to apply Graham's rule to subblock processing. As a result, the subject-matter of claim 1 is inventive starting from D2.

D3, not D2, should be regarded as the closest prior art. However, each individual pixel of D3 cannot be regarded as a block of size 1x1. The method of claim 1 is thus also inventive starting from D3.

D7 appears to have been submitted for confidential consideration at the July 1996 Tampere MPEG meeting. D7 is dated "July 1996". As explained in documents D14 and D15, at least some of the documents of the MPEG working group are restricted to MPEG members or might be post-published. Neither the examining division nor the board has provided evidence of the date on which D7 became available to the public. Thus D7 should be excluded from further consideration.

Reasons for the Decision

1. The appeal is admissible.

2. Entitlement of the claims to priority from P1

2.1 Amended claim 1 specifies the essential features of the adaptive DC prediction in a method of decoding MPEG-4 video signals. It expressly sets out that the DC coefficients DCA, DCB and DCC used in the steps of determining the first and second gradients are quantised, with blocks A, B and C being inside a boundary of a video object plane containing block X. The coefficient DCX of claim 1 is predicted according to the formulas in sections 2.1 and 2.2 of P1, and the DC coefficient is generated from the predicted coefficient DCX and the residual signal (see also Figures 1(b) and 2(b) of P1). The amendment thus has a direct basis in P1.

2.2 The examining division had objected to the coefficient DCX being predicted "based on" either coefficient DCC or coefficient DCA. Amended claim 1 sets out a step of predicting DCX "from" either DCC or DCA. This constitutes in the board's view a wording derivable from, and covering, both alternatives disclosed in the formulas in sections 2.1 and 2.2 of P1, read together with the passage (page 2, lines 6 to 8) mentioning prediction "from the horizontally adjacent or from vertically adjacent previous block". The amendment thus has a direct basis in P1.

2.3 Thus the board is convinced that the objections in the decision under appeal regarding the priority claimed from P1 do not apply to amended claim 1. The board is satisfied that the subject-matter of amended claim 1 is directly and unambiguously derivable from P1.

2.4 Independent apparatus claim 5 was amended accordingly, in particular with the coefficients DCA, DCB and DCC being quantised and with DCX being predicted from either DCC or DCA.

2.5 The board is also satisfied that the features of dependent claims 2 to 4, 6 and 7 are directly and unambiguously derivable from P1.

2.6 In conclusion, the subject-matter according to claims 1 to 7 and the subject-matter according to the earliest priority document P1 relate to "the same invention" within the meaning of Article 87(1) EPC 1973. Those claims thus enjoy the right to priority from P1.

3. Novelty (Article 54(1) EPC 1973)

3.1 Since the claims are entitled to priority from P1, they enjoy a date of priority, within the meaning of Article 89 EPC 1973, of 20 September 1996. Since document D1 was published on 12 February 1997, it is not comprised in the state of the art under Article 54(2) EPC 1973.

3.2 The examining division commented as an obiter dictum in the decision under appeal that European patent application published as EP 0 843 484 A1 might be relevant as prior art under Article 54(3) EPC. That document mentions international standards such as MPEG-1, H.261, MPEG-2 and H.263. However, it does not relate to MPEG-4 video signals, as set out in the present claims, so that its content is not novelty-destroying.

3.3 No further objection was raised under Article 54(1) EPC 1973 in the first-instance proceedings. The board is also satisfied that the subject-matter of the claims is novel over the further prior-art documents on file.

4. Inventive step (Article 56 EPC 1973)

4.1 Document D2 is in the board's view closer to the invention than D3 since D2 relates to a hybrid DCT/DPCM decoding method as the method of claim 1. Each 4x12 block in the frequency (DCT) domain is divided into three 4x4 subblocks (see Figures 3 and 4). The DC coefficient associated with a subblock X is predicted based on a comparison with coefficients of previously decoded subblocks neighbouring subblock X. Contrary to the present invention, D2 requires the transmission of overhead information ("directional bits") together with the video signals in order to determine which of the neighbouring subblocks is to be used for prediction.

4.2 The method of D2 and the method of claim 1 differ (inter alia) by the prediction scheme for the DC coefficient. The objective technical problem may thus be generally formulated as providing an alternative prediction scheme for a block-based decoding method.

4.3 Graham's rule according to D3 exploits the expected correlation between pixels, in particular in small (contiguous) picture regions, to predict the value of a picture element, thereby achieving compression without overhead transmission (see D3, column 1, lines 28 to 37 and column 5, lines 1 to 10). Further documents also refer to Graham's rule in the spatial (pel) domain (see D6, section "B. Adaptive Prediction" on pages 533 and 534; D8, section "Nonlinear Predictor" on pages 149 and 150; and D9, Figure 28 and the corresponding description on page 382).

4.4 On the one hand, orthogonal transformation into the frequency domain (for instance by discrete Fourier transform DFT in D11; or by discrete cosine transform DCT in D13 or in the MPEG-4 standard) aims to decorrelate transform coefficients within a block (see D11, the second paragraph of section 8.2.8 on page 566; and see D13, the first paragraph on page 29). On the other hand, prediction following Graham's rule essentially relies on correlation between neighbouring blocks (see D2, page 544, left-hand column, first paragraph; and see D10, section "PREDICTOR OPERATION PRINCIPLE" on page 6.1-1).

4.5 The DC coefficient in the frequency domain, which corresponds to the average intensity of a subblock (see D2, page 542, right-hand column, last paragraph), does not directly correspond to a spatially adjacent block of pixels in D2 as in the present invention, where adjacent 64x64 blocks in the frequency domain directly correspond to spatially adjacent blocks of 64x64 pixels, and where a DC coefficient thus directly corresponds to the average intensity of the corresponding adjacent blocks of pixels. Consequently, D2, by dividing blocks into subblocks prior to prediction, blurs the correspondence between DC coefficients in the frequency domain and corresponding adjacent picture regions, or their average intensity, in the spatial domain.

4.6 As a result, the board is not convinced that the skilled person would have readily envisaged replacing the DC prediction method according to D2 by the alternative method following Graham's rule known exclusively in the pixel domain. Consequently, the subject-matter of claim 1 is found to involve an inventive step over a combination of D2 with any one of documents D3, D6, D8 or D9.

4.7 Document D7 relates to the proposal by the working group (ISO/IEC JTC1/SC29/WG11) responsible for the MPEG-4 standard. D7 is dated "July 1996", which is the date of the Tampere meeting of the working group. According to D14, such a working group is usually a large group of experts from companies and organisations. However, the board cannot determine without further investigations whether the members of the particular working group for the MPEG-4 standard can be regarded as members of the public or whether they were bound to confidentiality. According to D14 and D15 (section "2 Media"), not all documents are published during a meeting, and some documents might be published after the meeting. From the evidence available, the board cannot determine whether D7 was made public during that meeting or at a later date preceding the earliest priority date of the present application, i.e. 20 September 1996. As a result, since there is insufficient evidence on file for the public availability of D7 before the relevant date of the present application, the board disregards D7 in the present appeal proceedings.

5. Further prosecution

In view of the above, the board considers that the reasons for refusing the present application, even taking account of the further comments made by the examining division, have been overcome by the amended claims, at least in so far as the cited documents belong to the state of the art under Article 54(2) EPC 1973 or Article 54(3) EPC. However, the issue of methods for decoding MPEG-4 signals has arisen for the first time in the appeal proceedings because of the amended claims. Since this issue was not examined by the examining division, the board remits the case to it for further prosecution.

Order

For these reasons it is decided that:

1. The decision under appeal is set aside.

2. The case is remitted to the department of first instance for further prosecution.