JVET-AL0216 [AHG9] Lens Optical Correction SEI – floating point parameters representation [G. Teniou, S. Wenger, A. Hinds (Tencent)]

This contribution provides justification on the choice of 25 bits representation to signal distortion parameters and suggests alternate approaches to signal the focal parameters.

The contribution includes the following:

1. How to signal the focal center (proposed syntax and semantics in clause 2.2.1)
2. How to signal the focal length (proposed syntax and semantics in clause 2.2.2.1 for option 1 or 2.2.2.2 for option 2.
3. Validate the signalling of radial distortion parameters (using i(25))
4. Agree the typo on the default value of one parameter in clause 2.4
5. Validate the signalling of chromatic aberration parameters (using u(25) and i(25))
6. Validate the signalling of vignetting parameters (using i(25))

No action was taken on the focal center proposal in section 2.2.1.

Two options proposed regarding focal length in section 2.2.2.

Curent design in TuC in JVET-AK2032:

loc_focal_length defines the focal length of the lens, i.e., the distance from the lens in which rays of light that fall parallel to the optical axis into the lens are converged. It is expressed in units of 0.1 millimeters. A value of 0 indicates that the focal length is undefined. When not present, the value of loc_focal_length is inferred to be equal to 0.

Option 1 of current contribution:

loc_focal_length defines the focal length of the lens, i.e., the distance from the lens in which rays of light that fall parallel to the optical axis into the lens are converged. It is expressed in units of millimeters. A value of 0 indicates that the focal length is undefined. When not present, the value of loc_focal_length is inferred to be equal to 0.

The current contribution removes the 0.1 millimeters to express in units of millimeters. The contribution states the following:

• 11 bits are needed to signal the integer part.
• 10 bits are necessary to signal 3 digits of the decimal part (10xln(2)/ln(10) = 3,01029)

However, there is no description in the semantics of how to convert the local_focal_length syntax element into a floating point number.

No action on focal length.

Section 2.3 provides a rationale for the current design of radial distortion and does not ask for a change to the design.

Section 2.4 asserts that there a bug in the inference value of loc_rd_poly4_k1 and proposes changing it 2²⁴, which corresponds to a value of 1, as opposed to the current value of 0.

Decision: agreed to chance the inference value of loc_rd_poly4_k1.

Section 2.5 provides a rationale for the current design of Transversal chromatic aberration and does not ask for a change to the design.

Section 2.6 provides a rationale for the current design of vignetting and does not ask for a change to the design.