diff --git a/08.decoding.process.md b/08.decoding.process.md index fccb8d09..048f3fd8 100644 --- a/08.decoding.process.md +++ b/08.decoding.process.md @@ -4408,20 +4408,6 @@ CurrFrame. The reconstruction and dequantization process is defined as follows: -The variable transpose is set equal to ( Tx_Width[ txSz ] > Tx_Height[ txSz ] ). - -The variable txSize is derived as follows: - - * If transpose is equal to 1, txSize is set equal to Transpose_Tx_Size[ txSz ]. - - * Otherwise (transpose is equal to 0), txSize is set equal to txSz. - -The variable txType is derived as follows: - - * If transpose is equal to 1, txType is set equal to Transpose_Tx_Type[ PlaneTxType ]. - - * Otherwise (transpose is equal to 0), txType is set equal to PlaneTxType. - The variable dqDenom is derived as follows: * If txSz is equal to TX_32X32, TX_16X32, TX_32X16, TX_16X64, or TX_64X16, dqDenom is set equal to 2. @@ -4446,11 +4432,11 @@ The variable tw is set equal to Min( 32, w ). The variable th is set equal to Min( 32, h ). -The variable flipUD is derived as follows. If txType is equal to one of FLIPADST_DCT, +The variable flipUD is derived as follows. If PlaneTxType is equal to one of FLIPADST_DCT, FLIPADST_ADST, V_FLIPADST, or FLIPADST_FLIPADST, flipUD is set equal to 1. Otherwise, flipUD is set equal to 0. -The variable flipLR is derived as follows. If txType is equal to one of DCT_FLIPADST, +The variable flipLR is derived as follows. If PlaneTxType is equal to one of DCT_FLIPADST, ADST_FLIPADST, H_FLIPADST, or FLIPADST_FLIPADST, flipLR is set equal to 1. Otherwise, flipLR is set equal to 0. @@ -4480,7 +4466,7 @@ The following ordered steps apply: f. Dequant[ i ][ j ] is set equal to Clip3( - ( 1 \<\< ( 7 + BitDepth ) ), ( 1 \<\< ( 7 + BitDepth ) ) - 1, dq2 ). 2. Invoke the 2D inverse transform block process defined in [section 7.12.2][] - with the variables txSize, txType and transpose as input. The inverse transform outputs are stored + with the variable txSz as input. The inverse transform outputs are stored in the Residual buffer. 3. For i = 0..(h-1), for j = 0..(w-1), the following applies: @@ -4490,7 +4476,7 @@ The following ordered steps apply: * The variable yy is set equal to flipUD ? ( h - i - 1 ) : i. * CurrFrame[ plane ][ y + yy ][ x + xx ] is set equal to - Clip1( CurrFrame[ plane ][ y + yy ][ x + xx ] + ( transpose ? Residual[ j ][ i ] : Residual[ i ][ j ] ) ). + Clip1( CurrFrame[ plane ][ y + yy ][ x + xx ] + Residual[ i ][ j ] ). It is a requirement of bitstream conformance that the values written into the Residual array in step 2 are representable by a signed integer with 8 + BitDepth bits. @@ -5042,13 +5028,7 @@ The process to invoke depends on n as follows: This process performs a 2D inverse transform for an array of coefficients stored in the 2D array Dequant. The output is placed in the 2D array Residual. -The inputs to this process are: - - * a variable txSz that specifies the transform size, - - * a variable txType that specifies the transform type, - - * a variable transpose, which, if equal to 1, signifies that the transform should be transposed. +The input to this process is a variable txSz that specifies the transform size. Set the variable log2W equal to Tx_Width_Log2[ txSz ]. @@ -5070,8 +5050,7 @@ The row transforms with i = 0..(h-1) are applied as follows: * T[ j ] is derived as follows for j = 0..(w-1): - * If i and j are both less than 32, T[ j ] is set equal to ( transpose ? Dequant[ j ][ i ] - \: Dequant[ i ][ j ] ). + * If i and j are both less than 32, T[ j ] is set equal to Dequant[ i ][ j ]. * Otherwise, T[ j ] is set equal to 0. @@ -5081,11 +5060,11 @@ The row transforms with i = 0..(h-1) are applied as follows: * If Lossless is equal to 1, invoke the Inverse WHT process as specified in [section 7.12.1.10][] with shift equal to 2. - * Otherwise, if txType is equal to one of DCT_DCT, ADST_DCT, FLIPADST_DCT or H_DCT, + * Otherwise, if PlaneTxType is equal to one of DCT_DCT, ADST_DCT, FLIPADST_DCT or H_DCT, invoke the inverse DCT process as specified in [section 7.12.1.3][] with the input variable n equal to log2W and the input variable r equal to rowClampRange. - * Otherwise, if txType is equal to one of DCT_ADST, ADST_ADST, DCT_FLIPADST, FLIPADST_FLIPADST, + * Otherwise, if PlaneTxType is equal to one of DCT_ADST, ADST_ADST, DCT_FLIPADST, FLIPADST_FLIPADST, ADST_FLIPADST, FLIPADST_ADST, H_ADST, or H_FLIPADST, invoke the inverse ADST process as specified in [section 7.12.1.9][] with input variable n equal to log2W and the input variable r equal to rowClampRange. @@ -5105,12 +5084,12 @@ The column transforms with j = 0..(w-1) are applied as follows: * If Lossless is equal to 1, invoke the Inverse WHT process as specified in [section 7.12.1.10][] with shift equal to 0. - * Otherwise, if txType is equal to one of DCT_DCT, DCT_ADST, DCT_FLIPADST or + * Otherwise, if PlaneTxType is equal to one of DCT_DCT, DCT_ADST, DCT_FLIPADST or V_DCT, invoke the inverse DCT process as specified in [section 7.12.1.3][] with the input variable n equal to log2H and the input variable r equal to colClampRange. - * Otherwise, if txType is equal to one of ADST_DCT, ADST_ADST, FLIPADST_DCT, + * Otherwise, if PlaneTxType is equal to one of ADST_DCT, ADST_ADST, FLIPADST_DCT, FLIPADST_FLIPADST, ADST_FLIPADST, FLIPADST_ADST, V_ADST, or V_FLIPADST, invoke the inverse ADST process as specified in [section 7.12.1.9][] with input variable n equal to log2H and the input variable r equal to diff --git a/docs/index.html b/docs/index.html index 79e5a9d0..0b3ec9d2 100644 --- a/docs/index.html +++ b/docs/index.html @@ -37,7 +37,7 @@

AV1 Bitstream & Decoding Process Specification

-

Last modified: 2018-04-05 04:05 PT

+

Last modified: 2018-04-05 08:44 PT

Authors
Peter de Rivaz, Argon Design Ltd
@@ -312,7 +312,10 @@

Contents

  • Large Scale Tile Decoding Process
  • Decode Frame Process
  • Ordering of OBUs
  • -
  • Random Access Decoding
  • +
  • Random Access Decoding +
  • CDF Update Process
  • Set Frame Refs Process
  • Motion Field Estimation Process
  • @@ -27034,19 +27171,7 @@

    2D Inverse Transform

    This process performs a 2D inverse transform for an array of coefficients stored in the 2D array Dequant. The output is placed in the 2D array Residual.

    -

    The inputs to this process are:

    - - +

    The input to this process is a variable txSz that specifies the transform size.

    Set the variable log2W equal to Tx_Width_Log2[ txSz ].

    @@ -27072,8 +27197,7 @@

    2D Inverse Transform