DETAILED ACTION
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/12/2026 has been entered.
Response to Arguments
Examiner incorporates here previous Responses to Arguments.
In view of the amendments to claim 8, the claim objection is withdrawn. Remarks, 9.
On pages 10–11 of the Remarks, Applicant contends the Sole Rojals and Coban are deficient for failing to teach or suggest the features added by way of amendment. Examiner finds the arguments moot in view of the new grounds of rejection necessitated by amendment. Specifically, the rejection of claim 1 now additionally relies on the combination of teachings of Sole Rojals and Coban, with the further teachings of Koo, Zhang, and Yoo. See claim rejections under 35 U.S.C. 103, infra.
On page 11 of the Remarks, Applicant contends forcing a coding unit-level index into a coefficient scan order is not workable. First, it is unclear how Applicant’s invention does not suffer the same fate. Second, the argument appears to fundamentally misapprehend the reason way Sole Rojals is cited. Sole Rojals is simply relied upon for teaching that the skilled artisan had in their possession the knowledge of transform coefficient syntax being context coded wherein the context can be based on previous transform coefficient information. The casual neighborhood concept articulated in Sole Rojals is well-known in the art and is simply a realization that context modeling based on previously coded symbols (temporally or spatially) adjacent to the current symbol being coded is obvious. Applicant’s argument that continuity is required is simply untrue and overlooks the fact that multiple contexts can be kept track of during encoding such that the encoder can address a second type of syntax element using a second context and get back to coding a first type of syntax element according to a first context model. Continuity may be a feature, but is not a requirement. Furthermore, Applicant makes a conclusory “breaking” argument without any evidence or reasoning. Attorney arguments and conclusory statements unsupported by factual evidence are entitled to little probative value. In re Geisler, 116 F.3d 1465, 1470 (Fed. Cir. 1997); see also In re Pearson, 494 F.2d 1399, 1405 (CCPA 1974) (indicating attorney argument is not evidence). For all the foregoing reasons, Examiner is unpersuaded of error.
Other claims are not argued separately. Remarks, 11.
35 USC § 112(f)
35 U.S.C. 112(f) reads as follows:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
Claim 20 discloses limitations that invoke 35 U.S.C. 112(f) under the analysis described in MPEP 2181.
According to MPEP 2181, 35 U.S.C. 112(f) is invoked by claim limitations that meet the following conditions: (1) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function; (2) the non-structural term is modified by functional language, typically, but not always linked by the transition word “for” or another linking word or phrase, such as “configured to” or “so that”; and (3) the non-structural term is not modified by sufficient structure, material, or acts for achieving the specified function. “Where a claim limitation meets the 3-prong analysis and is being treated under 35 U.S.C. 112, sixth paragraph, the examiner will include a statement in the Office action that the claim limitation is being treated under 35 U.S.C. 112, sixth paragraph.” MPEP 2181(I)(C).
In claim 20, Applicant uses the phrase “means for” for several limitations. In each case, Examiner interprets such language as a non-structural term followed by a linking word or phrase, which links the non-structural term to recited functions. In each case, the non-structural term is modified by functional language and is not modified by sufficient structure. Therefore, the claims invoke 35 U.S.C. 112(f). MPEP 2181(I).
Examiner finds that the units operable to perform the recited functions of claim 20 could be broadly construed as software modules or subroutines, or a computer or similar processing circuit or system of processors. Regarding the interpretation that the units might be a processor or group of processors, the Examiner considered whether the functions recited in claim 20 are functions typically found in a commercially available off-the-shelf processor. See In re Katz Interactive Call Processing Patent Litigation, 639 F.3d 1303, 1316 (Fed. Cir. 2011) (functions such as "processing," "receiving," and "storing" that can be achieved by any general purpose computer without special programming do not require disclosure of more structure than the general purpose processor that performs those functions). Because the recited functions are the substance and focus of the invention claimed, the Examiner finds these functions are not typically available in an off-the-shelf processor. Therefore, even if the means for terms, refer to processors (or similar), the Examiner finds the functional recitations do not connote to the skilled artisan sufficient structure of the processors (or similar) claimed. Accordingly, the phrase “means for” is interpreted as invoking the application of 35 U.S.C. § 112(f).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102 of this title, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 4–6, 8, and 11–20 are rejected under 35 U.S.C. 103 as being unpatentable over Sole Rojals (US 2013/0230097 A1), Coban (US 2012/0099646 A1), Koo (US 2023/0108690 A1), Zhang (US 2022/0417529 A1), and Yoo (US 2022/0201280 A1).
Examiner interprets the instant application to be drawn to the purported “inventive” concepts described in Applicant’s JVET-AD0204-v3 document. In that document it is explained that, “In order to model the context for a currently parsed coefficient when LFNST/NSPT is applied, the previous 5 coefficients in the coding order are used shown in Figure 1(right), rather than using the causal 2D neighborhood of context derivation.” Examiner notes Applicant described this approach in 2013, some 10 years before Applicant’s priority date. Sole Rojals (US 2013/0230097 A1), para. [0165] teaches the 5 coefficient “support” can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports. Zhang (US 2022/0417529 A1), ¶ 0236 also teaches either 5 coefficients in a causal template or in a template based on scan order. The other part of the purported “invention” is a general housekeeping item wherein the parsing of the lfnstIdx is moved earlier to indicate parsing the transform coefficients. “In ECM, the lfnstIdx is signalled after all the coefficients in a CU. For this proposal, lfnstIdx is signaled after all last_sig_coeff_pos syntax elements in a CU since lfnstIdx is required for parsing the transform coefficients.” JVET-AD0204-v3, Section 2.
Regarding claim 1, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests a method of decoding video data, the method comprising: decoding a low-frequency non-separable transform index (lfnstIdx) value for a coding unit including a current block of video data after parsing all last significant coefficient position syntax elements in the CU (Koo, ¶¶ 0399, 0421, and 0423: teaches the LFNST index can be signaled before signaling the transform coefficient information and after the last significant coefficient information); using the lfnstIdx value to select a context table from a plurality of context tables (Yoo, ¶ 0187: teaches each syntax element can have a context model associated therewith and can be implemented using a context model table, ctxTable); determining a context for context-based decoding a current transform coefficient of the current block of video data using the context table and according to previously coded transform coefficients (Sole Rojals, ¶ 0006: teaches, “Context for a bin of a syntax element may be determined based on values of related bins of previously coded syntax elements, such as syntax elements associated with other transform coefficients”), the previously coded transform coefficients being a sequence of immediately preceding transform coefficients to the current transform coefficient in scan order (Sole Rojals, ¶ 0006: teaches, “The locations from which context is derived may be referred to as a context derivation neighborhood (also referred to as "context support neighborhood" or simply "support").”; Sole Rojals, ¶ 0007: teaches, “Aspects of this disclosure generally relate to a context derivation neighborhood that is based on a transform coefficient scan order. For example, aspects of this disclosure relate to determining a support based on the order in which transform coefficients are scanned to serialize a two-dimensional array of transform coefficients to a one-dimensional array of transform coefficients (at a video encoder) or inverse scanned to reconstruct a two-dimensional array of transform coefficients from a one-dimensional array of transform coefficients (at a video decoder).”; Sole Rojals, ¶ 0165: teaches the 5 coefficient “support” can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports), the previously coded transform coefficients including the lfnstIdx value (Examiner notes that when LFNST is active, i.e. LFNST index > 0, coefficient level context modeling using templates may be disabled; Zhang, ¶¶ 0240 and 0241: teaches that when LFNST (RST) is enabled, then other syntax elements related to coefficient level passes are not signaled or are signaled using different context modeling; Zhang, ¶ 0236: teaches that when LFNST (RST) is enabled, then context modeling may be modified/determined separately from when it is not enabled; see also Zhang, ¶¶ 0256 and 0260: teaching different context modeling based on whether LFNST is enabled); context-based decoding the current transform coefficient using the determined context (Sole Rojals, ¶¶ 0006 and 0165: teach the transform coefficients are coded using the determined context) to form a one-dimensional array including M decoded transform coefficients where M is an integer greater than 0; inverse transforming the decoded transform coefficients using an MxN Karhunen-Loeve transform (KLT) to form a one-dimensional array including N residual values where N is an integer greater than 0 (Coban, ¶ 0115: teaches coefficient scanning serializes the “transform coefficients from a two-dimensional matrix to a one-dimensional array.”; Coban, ¶ 0115: teaches the transform can be a KLT, and as evidenced by a number of references cited under the Conclusion of this Office Action, when the prior art mentions KLT, the skilled artisan immediately understands KLT can be a 1-D KLT; see e.g. (1) Peringassery Krishnan; (2) Lan; (3) Rapaka; and/or (4) Zhao); and constructing a two-dimensional residual block including the residual values (Sole Rojals, ¶ 0007: teaches inverse scanning at the video decoder to form a two-dimensional array (i.e. block) of transform coefficients).
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals, with those of Coban, because both references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding would have been led to their relevant teachings, because, as skilled artisan knows (as evidenced by Choi, cited under Conclusion Section of Office Action explains), non-separable primary transforms, such as KLT, are better at removing correlation at the expense of greater complexity, and because Coban teaches that transforms other than Sole Rojal’s DCT transform, like KLT, is a well-known alternative used in the art such that the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals and Coban used in this Office Action unless otherwise noted.
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals and Coban, with those of Koo, because all three references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding would have been led to their relevant teachings and because, as Koo explains, the order of the LFNST index signaling being before or after transform coefficients is arbitrary and can be coded either before or after. Therefore, the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals, Coban, and Koo used in this Office Action unless otherwise noted.
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals, Coban, and Koo, with those of Zhang, because all four references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding would have been led to their relevant teachings and because, as Zhang explains, LFNST can be considered incompatible with utilizing certain context model templates or even just coding sig_coeff_flags, abs_level_gtx_flags, etc. Therefore, the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals, Coban, Koo, and Zhang used in this Office Action unless otherwise noted.
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals, Coban, Koo, and Zhang, with those of Yoo, because all five references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding using context models would have been led to their relevant teachings, because both Zhang and Yoo teach context modeling for certain syntax elements can differ according to context, like transform type, and because Yoo merely explains what the skill artisan already knew regarding how the art utilizes context tables for initializing context models for various syntax elements. Therefore, the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals, Coban, Koo, Zhang, and Yoo used in this Office Action unless otherwise noted.
Regarding claim 4, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 1, further comprising: using the lfnstIdx value to determine contexts for context-based coding values for one or more syntax elements of a corresponding transform coefficient comprising one of the previously coded transform coefficients (Examiner notes that when LFNST is active, i.e. LFNST index > 0, coefficient level context modeling using templates may be disabled; Zhang, ¶¶ 0240 and 0241: teaches that when LFNST (RST) is enabled, then other syntax elements related to coefficient level passes are not signaled or are signaled using different context modeling; Zhang, ¶ 0236: teaches that when LFNST (RST) is enabled, then context modeling may be modified/determined separately from when it is not enabled; see also Zhang, ¶¶ 0256 and 0260: teaching different context modeling based on whether LFNST is enabled).
Regarding claim 5, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 4, wherein the one or more syntax elements include one or more of a syntax element indicating whether the corresponding transform coefficient has an absolute value greater than zero, a syntax element indicating whether the corresponding transform coefficient has an absolute value greater than one, or a syntax element indicating whether the corresponding transform coefficient has an absolute value greater than two (Examiner notes the art calls these sig_coeff_flag, abs_level_gt1_flag, and abs_level_gt2_flag; Zhang, ¶¶ 0240–0241: teaches these flags can be signaled or skipped based on LFNST being enabled).
Regarding claim 6, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 4, wherein the one or more syntax elements include syntax elements for determining Rice parameter values used to parse remainder values for the corresponding transform coefficient (Examiner notes the art calls this syntax element the abs_remainder syntax element; Zhang, ¶¶ 0240–0241: teaches this syntax element can be signaled or skipped based on LFNST being enabled; Zhang, ¶ 0151: teaches the Rice parameter is used for coding the remainder syntax element).
Regarding claim 8, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 4, further comprising: determining that the lfnstIdx value is greater than zero; and in response to determining that the lfnstIdx value is greater than zero (Koo, ¶ 0381: teaches that when lfnst_idx is greater than 0, LFNST is applied): determining contexts for decoding a value of a significant coefficient flag syntax element, indicating whether the current transform coefficient has an absolute value greater than zero, using values of five of the previously coded transform coefficients, the five of the previously coded transform coefficients being coded in sequential order immediately prior to the current transform coefficient (Examiner notes that the art calls non-zero coefficients “significant coefficients”; Zhang, ¶¶ 0240–0241: teaches the sig_coeff_flag can be signaled when LFNST is enabled; Zhang, ¶ 0236: teaches the context model template can be the five neighboring previously coded information in scan order; Sole Rojals, ¶ 0165: teaches the 5 coefficient “support” can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports); determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than one using the values of the five of the previously coded transform coefficients; or determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than two using the values of the five of the previously coded transform coefficients (Examiner notes the art calls these abs_level_gt1_flag and abs_level_gt2_flag; Zhang, ¶¶ 0240–0241: teaches these flags can be signaled or skipped based on LFNST being enabled and subjected to context modeling as elsewhere disclosed; Zhang, ¶ 0236: teaches the context model template can be the five neighboring previously coded information in scan order; Sole Rojals, ¶ 0165: teaches the 5 coefficient “support” can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports).
Regarding claim 11, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 1, wherein using the lfnstIdx value to select the context table further comprises: determining a transform type for a transform used to transform the current transform coefficient (Zhang, ¶¶ 0232 and 0241: teaches context models can differ between transform types); and determining the context table corresponding to the transform type and using the lfnstIdx value (Yoo, ¶ 0187: teaches each syntax element can have a context model associated therewith and can be implemented using a context model table, ctxTable; Zhang, ¶¶ 0232 and 0241: teaches context models can differ between transform types).
Regarding claim 12, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 11, wherein determining the context table comprises: when the lfnstIdx value is equal to zero, determining the context table as being a regular coefficient context table; or when the lfnstIdx value is greater than zero, determining the context table as being a low-frequency non-separable transform coefficient context table (Examiner notes that lfnst_idx being zero means LFNST is not enabled for the current block; Zhang, ¶¶ 0232 and 0241: teaches context models for coefficients can differ depending on whether LFNST is enabled or not).
Regarding claim 13, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 12, wherein determining the context using the context table comprises determining the context according to contextTable[posX][posY][offset(sum(Abs(coeff)))], where contextTable comprises a two-dimensional matrix (Yoo, ¶ 0187: teaches each syntax element can have a context model associated therewith and can be implemented using a context model table, ctxTable), posX represents an X-position of the current transform coefficient in the current block, posY represents a Y-position of the current transform coefficient in the current block (Sole Rojals, ¶¶ 0031 and 0049: teaches the context model being the sum of the significant flags in the “support” is dependent on the location of the current transform coefficient), and offset(sum(Abs(coeff))) represents a sum of the absolute values of the previously coded transform coefficients offset according to an offset zone including the current transform coefficient (Sole Rojals, ¶¶ 0031 and 0049: teaches the context model being the sum of the significant flags in the “support”).
Regarding claim 14, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 11, wherein the current transform coefficient is included in an intra-prediction slice (I-slice) (Zhang, ¶ 0109: teaches reduced secondary transform, i.e. LFNST, is applicable to intra prediction (i.e. I-slices)).
Regarding claim 15, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 1, further comprising determining an offset to apply to an aggregate value of the previously coded transform coefficients according to a scan order position of the current transform coefficient (Zhang, ¶ 0236: teaches the scan order position is determined by applying an offset to the current coefficient, like -1, -2, …).
Regarding claim 16, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the method of claim 1, wherein determining the context comprises: determining a value of a low-frequency non-separable transform index; when the value of the low-frequency non-separable transform index is equal to zero, determining an offset according to a position of the current transform coefficient in the current block; when the value of the low-frequency non-separable transform index is greater than zero, determining the offset according to a scan order position of the current transform coefficient; and determining the context using the offset (Koo, ¶ 0381: teaches that when lfnst_idx is greater than 0, LFNST is applied; Zhang, ¶¶ 0240–0241: teaches these flags can be signaled or skipped based on LFNST being enabled and subjected to context modeling as elsewhere disclosed; Zhang, ¶ 0236: teaches the scan order position is determined by applying an offset to the current coefficient, like -1, -2, …; Sole Rojals, ¶ 0165: teaches the 5 coefficient “support” can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports).
Claim 17 lists the same elements as claim 1, but is drafted in apparatus form rather than method form. Therefore, the rationale for the rejection of claim 1 applies to the instant claim.
Regarding claim 18, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the device of claim 17, further comprising a display configured to display the decoded video data (Sole Rojals, ¶ 0056: teaches the coding techniques can be used in display devices).
Regarding claim 19, the combination of Sole Rojals, Coban, Koo, Zhang, and Yoo teaches or suggests the device of claim 17, wherein the device comprises one or more of a camera, a computer, a mobile device, a broadcast receiver device, or a set-top box (Sole Rojals, ¶ 0056: teaches the coding techniques can be used in the listed consumer electronics).
Claim 20 lists the same elements as claim 1, but is drafted in apparatus form rather than method form. Therefore, the rationale for the rejection of claim 1 applies to the instant claim.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Sole Rojals, Coban, Koo, Zhang, Yoo, and Haase (US 2021/0029360 A1).
Regarding claim 9, the combination of Sole Rojals, Coban, Koo, Zhang, Yoo, and Haase teaches or suggests the method of claim 4, further comprising: determining that the lfnstIdx value is greater than zero (Koo, ¶ 0381: teaches that when lfnst_idx is greater than 0, LFNST is applied); determining a size of the current block; determining a number of previously coded transform coefficients according to the size of the current block (Haase, ¶ 0010: teaches setting a shape of a local transform coefficient context template or disabling the local template based on size of the block, color component, position of last significant coefficient, transform type, use of context from previously encoded transform coefficients located at positions determined by the local template, etc.; see also Haase, ¶ 0211); and in response to determining that the lfnstIdx value is greater than zero (Koo, ¶ 0381: teaches that when lfnst_idx is greater than 0, LFNST is applied): determining contexts for decoding a value of a significant coefficient flag syntax element, indicating whether the current transform coefficient has an absolute value greater than zero, using values of the number of the previously coded transform coefficients, the number of the previously coded transform coefficients being coded in sequential order immediately prior to the current transform coefficient; determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than one using the values of the number of the previously coded transform coefficients; or determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than two using the values of the number of the previously coded transform coefficients (see treatment of original claim 8).
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals, Coban, Koo, Zhang, and Yoo, with those of Hasse, because all six references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding would have been led to their relevant teachings and because, as Haase explains, basing a context model template for transform coefficients on block size was known to the skilled artisan prior to Applicant’s priority date and is, thus, obvious in view of the teachings of the prior art. Therefore, the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals, Coban, Koo, Zhang, Yoo, and Haase used in this Office Action unless otherwise noted.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Sole Rojals, Coban, Koo, Zhang, Yoo, and Coban et al., “Algorithm description of Enhanced Compression Model 8 (ECM 8),” JVET-AC2025, 29th Meeting: by teleconference, January 2023 (herein “Coban-2”).
Regarding claim 10, the combination of Sole Rojals, Coban, Koo, Zhang, Yoo, and Coban-2 teaches or suggests the method of claim 1, further comprising: coding a non-separable primary transform index (nsptIdx) value for the current block after coding a last significant coefficient syntax element value for the current block (Coban-2, pg. 55, Section 3.3.6: teaches NSPT for intra coding wherein NSPT is similar to but replaces LFNST for certain block shapes and would have to have an index, like the lfnst_idx, indicating which transform in a transform set is selected; NSPT is just a specialized approach to LFNST for certain block sizes/shapes, the skilled artisan would find it arbitrary when to signal the index just like with LFNST and as taught by Koo; Koo, ¶¶ 0399, 0421, and 0423: teaches the LFNST index can be signaled before signaling the transform coefficient information and after the last significant coefficient information); when the nsptIdx value is greater than zero (As already stated, the nspt_Idx, because it is an extension really of LFNST for certain block shapes, would behave similarly to lfnst_Idx, especially given Coban-2’s teaching that NSPTs consist of 35 sets and 3 candidates similar to current LFNST; Koo, ¶ 0381: teaches that when lfnst_idx is greater than 0, LFNST is applied): determining contexts for decoding a value of a significant coefficient flag syntax element, indicating whether the current transform coefficient has an absolute value greater than zero, using values of a number of the previously coded transform coefficients, the number of the previously coded transform coefficients being coded in sequential order immediately prior to the current transform coefficient; determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than one using the values of the number of the previously coded transform coefficients; or determining contexts for decoding a value of a syntax element indicating whether the current transform coefficient has an absolute value greater than two using the values of the number of the previously coded transform coefficients (see treatment of original claim 8).
One of ordinary skill in the art, before the effective filing data of the claimed invention, would have been motivated to combine the elements taught by Sole Rojals, Coban, Koo, Zhang, and Yoo, with those of Coban-2, because all six references are drawn to the same field of endeavor such that one wishing to practice the art of transform coefficient coding would have been led to their relevant teachings and because, as Coban-2 explains, non-separable primary transforms (NSPT) replace, but are similar to, LFNSTs for certain block shapes, and thus, modifying Zhang’s treatment of LFNST context modeling to include context modeling for Coban-2’s NSPTs is obvious in view of the teachings of the prior art. Therefore, the combination is a mere combination or prior art elements, according to known methods, to yield a predictable result. This rationale applies to all combinations of Sole Rojals, Coban Koo, Zhang, Yoo, and Coban-2 used in this Office Action unless otherwise noted.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Haase (US 2021/0029360 A1) teaches setting a shape of a local transform coefficient context template or disabling the local template based on size of the block, color component, position of last significant coefficient, transform type, use of context from previously encoded transform coefficients located at positions determined by the local template, etc. (¶ 0010). Haase also teaches a local template being 5 previously coded coefficients in a certain neighboring shape, but that any number and location of template positions can be chosen (Fig. 10 and ¶ 0129). Haase also teaches the template can be used for any pass or using single-pass (¶ 0133). Secondary transforms (¶ 0194). Adaptive shape/size, for example, 5 or 3 (¶ 0211). Template position offset to current position (¶ 0040).
Karczewicz (US 2020/0077117 A1) teaches a context template for coefficient levels and sign (Fig. 4 and ¶ 0104). Also teaches ricePar and neighboring coefficients in the context template (¶ 0146).
Coban (US 2020/0007873 A1) teaches context coding each pass using a template rather than having to wait for the entirety of the previous coefficient’s information to be decoded before using it to predict the current context (¶¶ 0033–0037).
Sole Rojals (US 2013/0230097 A1) teaches a five position-based support for defining a context model for coding bins of a significance map wherein the probability model is “defined as a sum of the significant flags in every point of the support” (¶ 0031 and Fig. 5). The publication also teaches the support can be based on scan order so that the support locations substantially overlap between successive coefficient positions to “reduce the computational and data access expenses” of position-based supports. (¶ 0165).
Choi (US 2024/0357111 A1) teaches non-separable primary transforms are more efficient than separable primary transforms and that separable transforms utilize separate transform kernels for the horizontal and vertical directions (e.g. ¶ 0186).
Peringassery Krishnan (US 2023/0100043 A1) teaches transform kernels can be separable 1-D transform kernels or non-separable 2-D transform kernels and can be DCT, DST, KLT, etc. (¶ 0183).
Guleryuz (US 2020/0092553 A1) teaches that good compression performance can be achieved using a non-separable KLT rather than a separable DCT at the expense of computational complexity (¶ 0004).
Arrufat (US 2018/0359492 A1) teaches non-separable KLT is considered optimal in terms of data decorrelation and compression efficiency, but comes at a cost of higher complexity as compared to separable transforms (¶ 0041).
Nguyen et al., “Non-CE11: Proposed Cleanup for Transform Coefficient Coding,” JVTVC-H0228, 8th Meeting: San Jose, CA, Feb. 2012. The publication teaches the local template of five previous coefficients, single-pass vs. multi-pass, summing abs coefficient levels, bypass Rice coding, etc.
Zhao (US 2020/0389666 A1) teaches signaling LFNST index after last non-zero coefficient information, but before sig_coeff_flag, gtX flags, parity flags, sign flags, and remainder flags (¶ 0172).
Browne et al., “Algorithm description for Versatile Video Coding and Test Model 19 (VTM19),” JVET-AC2002-v1, 29th Meeting: Teleconference, January 2023. The publication teaches Transform and Quantization, including LFNST (Section 3.5, pg. 61, et seq.), Entropy Coding, including ctxTable and interval computation, transform coefficient level coding, Rice Parameter coding, etc. (Section 3.1, pg. 69, et seq.), and Context modeling for coefficient coding (Section 3.6.4, pg. 73, et seq.).
Lan et al., “An Improved JPEG Image Coder Using the Adaptive Fast Approximate Karhunen-Loeve Transform (AKLT),” 1994 International Symposium on Speech, Image Processing and Neural Networks, 13–16 April 1994, Hong Kong. This publication teaches a KLT applies to a digital image can have better performance than DCT (Section 1) and teaches the forward KLT and inverse KLT are one-dimensional operating on an input vector of length N (Section 2).
Rapaka (US 2014/0064360 A1) teaches 1-D KLT used for transforming the residual signal (¶ 0214).
Zhao (US 2023/0247209 A1) teaches 1-D transforms include KLT (e.g. ¶ 0136).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Michael J Hess whose telephone number is (571)270-7933. The examiner can normally be reached on Mon - Fri 9:00am-5:30pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, William Vaughn can be reached on (571)272-3922. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
MICHAEL J. HESS
Primary Examiner
Art Unit 2481
/MICHAEL J HESS/Primary Examiner, Art Unit 2481
Prosecution Insights