VTM is missing conformance check on TransCoeffLevel for dependent quantization.

This setting of TransCoeffLevel in the spec residual coding syntax for dependent quantization (sh_dep_quant_used_flag = 1):

TransCoeffLevel[ x0 ][ y0 ][ cIdx ][ xC ][ yC ] =
( 2 * AbsLevel[ xC ][ yC ] − ( QState > 1 ? 1 : 0 ) ) *
( 1 − 2 * coeff_sign_flag[ n ] )

is implemented in line 777 of DepQuant.cpp:

Intermediate_Int qIdx = 2 * level + (level > 0 ? -(state >> 1) : (state >> 1));

TransCoeffLevel must be in the range of CoeffMin to CoeffMax, inclusive. So, a check on qIdx should be added after line 777 to catch non conformance bitstreams.

In the semantics for ph_recovery_poc_cnt, the following correction should be made:

If the current picture is a GDR picture and ph_recovery_poc_cnt is equal to 0, the current picture itself is also referred to as the recovery point point picture.

In the last part in section 8.7.5.1, IsAvailable is set to use for checking whether a neighboring block is available in next CU decoding as shown below.

The following assignments are made for i = 0..nCurrSw * SubWidthC − 1, j = 0..nCurrSh* SubHeightC − 1:

IsAvailable[ cIdx ][ xCurr * SubWidthC + i ][ yCurr * SubHeightC + j ] = TRUE (1201)

When cIdx == 0 (luma case), the coordinate calculation seems to have an issue. It seems SubWidthC and SubHeightC should not be used for luma, but only used for chroma. I asked an expert about this issue, and he confirmed the issue and kindly provided the following fix. Please take a look at this and correct me know if I am wrong.

The variables subW and subH are derived as follows:

subW = cIdx = = 0 ? 1 : SubWidthC (xxxx) subH = cIdx = = 0 ? 1 : SubHeightC (xxxx)

The following assignments are made for i = 0..nCurrSw * subW − 1, j = 0..nCurrSh * subH − 1:

IsAvailable[ cIdx ][ xCurr * subW + i ][ yCurr * subH + j ] = TRUE (1201)