NEC Corporation filed Patent Application No. 201717010986 before the Indian Patent Office in March 2017, a national phase entry from PCT Application PCT\/JP2015\/004037, claiming priority from a Japanese application of October 2014. The invention relates to a video coding device combining two techniques: adaptive colour transform in residual domain (choosing, per coding unit, whether to compress the prediction error in the RGB or YCoCr colour space) and a chroma quantization offset (adjusting quantization strength per colour component to improve subjective image quality). The technical problem NEC identified was this: when both techniques are used together, blocks in RGB space and blocks in YCoCr space share the same quantization strength, eliminating the subjective quality benefit of the chroma quantization technique.<\/p>\n
After a First Examination Report in January 2020 and a hearing in April 2021, the Assistant Controller of Patents and Designs refused the application on 21 June 2022 under Section 15 of the Patents Act for lack of inventive step. The three prior art documents cited were: HEVC Screen Content Coding Draft Text 1 (‘D1’), a paper on in-loop colour-space transform (‘D2’), and PCT publication WO2015143671 (‘D3’). NEC appealed under Section 117A(2) of the Patents Act.<\/p>\n
The court found that D1, when Sections 8.6.2, 8.4, and 8.6.8 are read together, discloses selecting among a plurality of colour spaces at the coding unit level. Section 8.6.2 sets out equations through which the quantization parameter for each colour component is derived by applying an offset of -5, -3, or 0 depending on the adaptive colour transform flag: a flag value of 1 indicates the transform has been applied to the current coding unit. The combination of these sections discloses per-colour-space QP adjustment with different offset values depending on which colour space is active. The court held this directly anticipates the adaptive chroma quantization offset derivation feature that NEC claimed as its inventive contribution.<\/p>\n
D2 explicitly discloses delta QPs of (-5, -3, -5) for YCoCg when colour space transform is applied. The court reasoned that a person skilled in the art would immediately understand that if the transform is not applied, the normal quantization parameter remains unmodified, implying a delta QP of 0 for the original colour space. This implicit understanding reinforces the prior art picture on per-colour-space adaptive offset derivation without requiring any creative step.<\/p>\n
On D3, the court acknowledged NEC’s objection that it was published after the priority date and its claims had not been granted. However, the court held that D3’s claims must be read in conjunction with its specification, and that paragraph [0149] of D3 discloses two separate QP syntax elements for the first and second colour space respectively, indicating that QP adjustment is applicable to both colour spaces. The court did not rely solely on D3, but treated it as corroborative support alongside D1 and D2.<\/p>\n
The court found NEC’s shifting positions damaging to its case. In post-hearing submissions before the Controller, NEC had accepted that D1 teaches derivation of an adaptive chroma quantization offset for each of the plurality of colour components. In the appeal, NEC argued that Section 8.6.2 describes no chroma quantization offset at all, and also argued that the section describes only an adjustment to the luma component. Neither of these positions was taken at the hearing stage. The court declined to accept arguments that contradicted concessions already made below, and held that in view of D1 and D2, the claimed subject matter was rendered obvious to a person skilled in the art. The appeal was dismissed.<\/p>\n