Method and device for image encoding/decoding using block split prediction
A video encoding/decoding method and apparatus using split prediction is provided. The video encoding/decoding method and apparatus using split prediction splits a block into various partition shapes in addition to a conventional standardized right-angled quadrangular shape so as to effectively predict an image of a current block from a previous image and thus, may more improve the performance of video data compression and may achieve superior reconstructed image quality.
1. A video encoding apparatus for encoding a current block which is located in a current frame, the apparatus comprising:
an inter-predictor configured to
generate a reference template surrounding at least two sides of the current block, the reference template comprised of pixels within one or more pre-encoded adjacent blocks in the current frame which are adjacent to the current block, wherein the reference template does not include pixels in the current block,
search for, in a previously encoded reference frame other than the current frame, a predicted template which has a shape corresponding to the reference template and pixel values similar to the reference template, and
set split information of the current block as split information of a central block surrounded by pixels in the predicted template, and generate a predicted block by predicting each partition divided from the current block based on the split information of the current block;
a subtractor configured to generate a residual block by subtracting the predicted block from the current block;
a transformer and quantizer configured to generate a transformed and quantized residual block by transforming and quantizing the residual block; and
an encoder configured to encode the transformed and quantized residual block, without encoding the split information of the current block.
2. The apparatus of claim 1 , wherein the split information corresponds to location information associated with a line that splits the central block.
3. The apparatus of claim 1 , wherein the split information corresponds to location information associated with a feature point of the central block.
4. A video decoding apparatus for decoding a current block which is located in a current frame, the apparatus comprising:
a decoder configured to reconstruct a transformed and quantized residual block by receiving encoded data, wherein split information of the current block is not reconstructed from the encoded data;
an inverse-quantizer and inverse-transformer configured to reconstruct a residual block by inverse-quantizing and inverse-transforming the transformed and quantized residual block;
an inter-predictor configured to extract the split information of the current block from one or more pre-decoded adjacent blocks of the current block, and generate a predicted block by predicting each partition divided from the current block based on the split information of the current block; and
an adder configured to reconstruct the current block by adding the reconstructed residual block and the predicted block,
wherein the inter-predictor is configured to
generate a reference template surrounding at least two sides of the current block, the reference template comprised of pixels within the one or more pre-decoded adjacent blocks in the current frame which are adjacent to the current block, wherein the reference template does not include pixels in the current block,
search for, in a previously decoded reference frame other than the current frame, a predicted template which has a shape corresponding to the reference template and pixel values similar to the reference template, and
set the split information of the current block as split information of a central block surrounded by pixels in the predicted template, and generate the predicted block by predicting each partition divided from the current block based on the split information of the current block.
5. The apparatus of claim 4 , wherein the split information corresponds to location information of a line that splits the central block.
6. The apparatus of claim 4 , wherein the split information corresponds to location information associated with a feature point of the central block.
7. A video decoding method for decoding a current block which is located in a current frame, the method comprising:
reconstructing a transformed and quantized residual block by receiving encoded data, wherein split information of the current block is not reconstructed from the encoded data;
reconstructing a residual block by inverse-quantizing and inverse-transforming the transformed and quantized residual block;
generating a predicted block by extracting the split information of the current block from one or more pre-decoded adjacent blocks divided from the current block, and by predicting each partition of the current block based on the split information of the current block; and
reconstructing the current block by adding the reconstructed residual block and the predicted block,
wherein the generating of the predicted block comprises:
generating a reference template surrounding at least two sides of the current block, the reference template comprised of pixels within the one or more pre-decoded adjacent blocks in the current frame which are adjacent to the current block, wherein the reference template does not include pixels in the current block;
searching for, in a previously decoded reference frame other than the current frame, a predicted template which has a shape corresponding to the reference template and pixel values similar to the reference template, and
setting the split information of the current block as split information of a central block surrounded by pixels in the predicted template, and generating the predicted block by predicting each partition of the current block based on the split information of the current block.
8. The method of claim 7 , wherein the split information corresponds to location information of a line that splits the central block.
9. The method of claim 7 , wherein the split information corresponds to location information associated with a feature point of the central block.