In 8.4.3.1.6, when intraPredMode is 18, Table 8-5 sets intraPredAngle to -32. Then, after equation (8-47) is applied, (8-48) is applied because intraPredAngle is less than 0. However, (8-48) uses invAngle, and Table 8-6 does not specify invAngle for intraPredMode=18.

Similarly, suppose intraPredMode is between 11 and 17, inclusive. This is less than 18, so (8-50) is applied. Because modes 11-17 have intraPredAngle values less than zero, (8-51) is applied. However, (8-51) uses invAngle, and Table 8-6 does not contain assign invAngle values to modes 11-17.

Modifying the top row of Table 8-6 may be a solution, by changing the values

From: 2, 3, 4, 5, 6, 7, 8, 9 To: 18, 17, 16, 15, 14, 13, 12, 11

We may also want to see if this affects ticket #339.

In chapter 8.4.4.2.2, the substitution is described as following :

"The values of the samples p[ x ][ y ] with x = −1, y = −1..nT*2−1 and x = 0..nT*2−1, y = −1 are modified as follows: – If all samples p[ x ][ y ] with x = −1, y = −1..nT*2−1 and x = 0..nT*2−1, y = −1 are marked as "not available for intra prediction," the value ( 1 << ( BitDepthY − 1 ) ) is substituted for the values of all samples p[ x ][ y ]."

This is applied for luma and chroma pixels. For chroma pixels, I think the value to be used should be : ( 1 << ( BitDepthC − 1 ) )

In chapter 8.4.4.2.5 of JCTVC-K1003_v4, the variable nS is used for the calculation of dcVal. nS is not defined in this chapter, but earlier as 1 << Log2CbSize. I think there is a typo there. nT (the transform block size) should be used instead of nS.