Method and device for coding residual signal in video coding system

1. An image decoding method performed by a decoding device, comprising:

deriving quantized transform coefficients by unit of sub-blocks in a transform unit based on residual information included in a bitstream;

deriving transform coefficients based on the quantized transform coefficients;

generating a residual sample based on the transform coefficients;

generating a prediction sample based on an inter prediction or an intra prediction; and

restoring an image based on the residual sample and the prediction sample,

wherein each sub-block included in a first group among the sub-blocks in the TU uses an up-right diagonal scan order, wherein the quantized transform coefficients of a sub-block included in the first group are decoded according to the up-right diagonal scan order from an upper left position to a lower right position,

wherein each sub-block included in a second group among the sub-blocks in the TU uses a reverse order of the up-right diagonal scan order, wherein the quantized transform coefficients of a sub-block included in the second group are decoded according to the reverse order of the up-right diagonal scan order from a lower right position to an upper left position, and

wherein the first group includes upper left sub-blocks in the TU.

2. The decoding method of claim 1 , further comprising:

dividing the TU into multiple sub-blocks,

wherein the sub-blocks have a size of 8×8 or larger.

3. The decoding method of claim 2 , wherein when the size of the TU is 64×64 or larger, the sub-blocks divided from the TU have a size of 8×8 or larger.

4. The decoding method of claim 1 , wherein the residual information includes the information indicating non-zero quantization transform coefficients of the sub-block and flag information indicating whether the absolute values of all of the non-zero quantization transform coefficients in the sub-block are 1, and

the deriving of the quantized transform coefficients by unit of sub-blocks includes,

detecting whether the absolute values of the quantization transform coefficients are 0 sequentially based on a specific scan order from the lower right position of the sub-block based on the information indicating the non-zero quantization transform coefficients, and

detecting whether the absolute values of all of the non-zero quantization transform coefficients in the sub-block are 1 based on the flag information.

5. The decoding method of claim 4 , wherein the residual information further includes information indicating signs of the non-zero quantization transform coefficients in the sub-block, and

the deriving of the quantized transform coefficients by unit of sub-blocks further includes

detecting the signs of the quantization transform coefficients larger than 0 based on information indicating the signs of the quantization transform coefficients larger than 0 when a value of the flag information is 1.

6. The decoding method of claim 1 , wherein the residual information includes the information indicating non-zero quantization transform coefficients of the sub-block and flag information indicating whether the absolute values of all non-zero AC component quantization transform coefficients in the sub-block are 1, the sub-block including DC component quantization transform coefficients, and

the deriving of the quantized transform coefficients by unit of sub-blocks includes

detecting whether the absolute values of the quantization transform coefficients are 0 sequentially based on a specific scan order from the lower right position of the sub-block based on the information indicating the non-zero quantization transform coefficients, and

detecting whether the absolute values of all of the non-zero AC component quantization transform coefficients excluding DC component quantization transform coefficient of the sub-block are 1 based on the flag information.

7. The decoding method of claim 6 , wherein the residual information further includes information indicating whether the absolute value of the DC component quantization transform coefficient is larger than 1, and

the deriving of the quantized transform coefficients by unit of sub-blocks includes

deriving the absolute value of the DC component quantization transform coefficient based on the information indicating whether the absolute value of the DC component quantization transform coefficient is larger than 1.

8. An image encoding method performed by an encoding device, comprising:

generating a prediction sample based on an inter prediction or an intra prediction;

generating a residual sample based on a comparison of an original sample and the prediction sample;

deriving transform coefficients based on the residual sample;

deriving quantized transform coefficients based on the transform coefficients; and

encoding and outputting residual information for the quantized transform coefficients by unit of sub-blocks in a transform unit (TU),

wherein each sub-block included in a first group among the sub-blocks in the TU uses an up-right diagonal scan order, wherein the quantized transform coefficients of a sub-block included in the first group are encoded according to the up-right diagonal scan order from an upper left position to a lower right position,

wherein each sub-block included in a second group among the sub-blocks in the TU uses a reverse order of the up-right diagonal scan order, wherein the quantized transform coefficients of a sub-block included in the second group are encoded according to the reverse order of the up-right diagonal scan order from a lower right position to an upper left position, and

wherein the first group includes upper left sub-blocks in the TU.