Method and device for encoding/decoding image by inter prediction using random block

1. A video encoding apparatus, comprising:

a block partitioning unit configured to determine a partition form, among candidate partition forms, for partitioning a current block into one or more partition blocks, wherein the candidate partition forms include one or more asymmetric partition forms in which the current block is partitioned into a plurality of asymmetric partition blocks;

a prediction unit configured to generate one or more predicted blocks respectively corresponding to the one or more partition blocks;

a subtractor configured to generate a residual block of the current block by subtracting the predicted blocks from one or more partition blocks corresponding thereto;

a transformer configured to generate at least one transform block by transforming the residual block;

a quantizer configured to generate at least one quantized transform block by quantizing the at least one transform block; and

an encoder configured to encode the at least one quantized transform block and information on the determined partition form into a bitstream,

wherein the prediction unit is configured to

determine a scale factor, and

predict a partition block partitioned from the current block, by deriving predicted pixels from a reference picture in a scale corresponding to the scale factor from a pixel location indicated by a motion vector of the partition block, and

wherein the encoder is configured to encode information on the scale factor into the bitstream to enable a video decoding apparatus to derive the predicted pixels from the reference picture in a scale corresponding to the information on the scale factor.

2. The video encoding apparatus of claim 1 , wherein the information on the determined partition form is encoded with reference to a partition form encoding table.

3. The video encoding apparatus of claim 1 , wherein the information on the scale factor is encoded with reference to a scale factor encoding table.

4. The video decoding apparatus of claim 1 , wherein the asymmetric partition forms include partition forms for partitioning the current block into two partition blocks, one of the two partition blocks having a rectangular shape in 1/4 size of the current block and the other block having a rectangular shape in 3/4 size of the current block.

5. A video decoding apparatus, comprising:

a decoder configured to reconstruct at least one quantized transform block, information on a partition form and information on a scale factor from the bitstream, and

a dequantizer configured to reconstruct at least one transform block by dequantizing the at least one quantized transform block;

an inverse transformer configured to reconstruct a residual block of a current block to be decoded, based on inversely transforming the at least one transform block;

a prediction unit configured to generate one or more partitioned prediction blocks of the current block partitioned according to a partition form indicated by the information on the partition form among candidate partition forms, wherein the candidate partition forms include one or more asymmetric partition forms in which the current block is partitioned into a plurality of asymmetric partition blocks; and

an adder configured to reconstruct the current block based on the reconstructed residual block and the prediction blocks,

wherein the prediction unit is configured to predict a partition block partitioned from the current block, by deriving predicted pixels from a reference picture in a scale corresponding to the information on the scale factor from a pixel location indicated by a motion vector of the partition block.

6. The video decoding method of claim 5 , wherein the asymmetric partition forms include partition forms for partition forms is for partitioning the current block into two partition blocks, one of the two partition blocks having a rectangular shape in 1/4 size of the current block and other block having a rectangular shape in 3/4 size of the current block.

7. A video decoding method performed by a video decoding apparatus, the method comprising:

reconstructing at least one quantized transform block, information on a partition form and information on a scale factor from the bitstream;

reconstructing at least one transform block by dequantizing the least one quantized transform block;

reconstructing a residual block of a current block to be decoded, based on inversely transforming the least one transform block;

generating a prediction block of the current block by predicting one or more partition blocks which are partitioned from the current block according to a partition form indicated by the information on the partition form among candidate partition forms, wherein the candidate partition forms include one or more asymmetric partition forms in which the current block is partitioned into a plurality of asymmetric partition blocks; and

reconstructing the current block based on the reconstructed residual block and the prediction block,

wherein the generating of one or more partitioned prediction blocks comprises predicting a partition block partitioned from the current block, by deriving predicted pixels from a reference picture in a scale corresponding to the information on the scale factor from a pixel location indicated by a motion vector of the partition block.