| Test B.5 – Escape colour prediction
Proponent: InterDigital, JCTVC-R0170, JCTVC-S0132
Crosschecker: ITRI, JCTVC-S0211
In this test, an escape colour prediction method was tested. A flag is signalled to indicate if all colour components of the current escape colour pixel are the same as that of the previous escape colour pixel. If not, the current escape colour value is predicted from an entry in the palette table. The index value of the table entry used for prediction and the prediction difference are coded.
Note: In addition to entries in the palette table, previous coded escape colours can also be used for predicting the current escape colour.
It was commented that this seems essentially similar to just using a larger palette or palette predictor.
Gain is very minor except for lossless case. Gain is significant for lossless case. However, the encoding time is substantially increased.
Further study in CE suggested. Having a faster encoding method would be desirable.
In later discussion chaired by GJS on Friday 10-24, it was said that the proposal introduces several (seven) syntax elements, several of which are context coded, and that it complicates the design. It was asserted that other more straightforward methods, such as increased palette size, could also improve compression for the lossless case. Participants other than the proponent expressed the view that this should not be tested in a CE.
Dostları ilə paylaş: | 
