Prediction image generation method, image coding method, image decoding method, and prediction image generation apparatus

Abstract

Provided is a prediction image generation method for generating a prediction image of a current block, the prediction image generation method including: an extraction step of extracting a plurality of first feature points each of which has a local feature quantity, the plurality of first feature points being included in a reconstructed image; a search step of searching a corresponding point from the plurality of first feature points, the corresponding point having a local feature quantity similar to a local feature quantity of a second feature point corresponding to the current block, a relationship between the corresponding point and the second feature point being expressed by information including a non-parallel translation component; and a generation step of generating the prediction image from the reconstructed image based on the relationship.

What is claimed is: 
     
         1 . A prediction image generation method for generating a prediction image of a current block, the prediction image generation method comprising:
 an extraction step of extracting a plurality of first feature points each of which has a local feature quantity, the plurality of first feature points being included in a reconstructed image;   a search step of searching a corresponding point from the plurality of first feature points, the corresponding point having a local feature quantity similar to a local feature quantity of a second feature point corresponding to the current block; and   a generation step of generating the prediction image from the reconstructed image based on a relationship between the corresponding point and the second feature point being expressed by information including a non-parallel translation component.   
     
     
         2 . The prediction image generation method according to  claim 1 , wherein the second feature point is included in the current block, and
 in the generation step, the prediction image is generated using a pixel value of a region including the corresponding point in the reconstructed image.   
     
     
         3 . The prediction image generation method according to  claim 1 , wherein the second feature point is a feature point in surroundings of the current block, and
 in the generation step, the prediction image is generated using a pixel value of a region that does not include the corresponding point in the reconstructed image.   
     
     
         4 . The prediction image generation method according to  claim 3 , wherein the reconstructed image is a reconstructed image of a current picture including the current block. 
     
     
         5 . The prediction image generation method according to  claim 3 , wherein the reconstructed image is a reconstructed image of a picture different from a current picture including the current block. 
     
     
         6 . An image coding method in which the prediction image generation method according to  claim 5  is performed, the image coding method comprising:
 an image coding step of coding the current block using the prediction image. 
 
     
     
         7 . The image coding method according to  claim 6 , further comprising:
 a feature point information coding step of coding feature point information identifying the second feature point in a plurality of third feature points corresponding to the current block,   wherein the plurality of third feature points are extracted in the extraction step, and   the second feature point is selected from the plurality of third feature points in the search step.   
     
     
         8 . The image coding method according to  claim 7 , wherein the feature point information indicates a coordinate of the second feature point. 
     
     
         9 . The image coding method according to  claim 7 , wherein the feature point information indicates a rotation quantity or a scale value which is possessed by the second feature point. 
     
     
         10 . The image coding method according to  claim 9 , further comprising:
 a corresponding point information coding step of coding corresponding point information identifying the corresponding point in the plurality of first feature points.   
     
     
         11 . The image coding method according to  claim 10 , wherein the corresponding point information indicates a coordinate of the second feature point. 
     
     
         12 . The image coding method according to  claim 10 , wherein, in the feature point information coding step,
 indexes are allocated to the plurality of first feature points in a predetermined sequence, and   the corresponding point information indicates the index allocated to the corresponding point.   
     
     
         13 . The image coding method according to  claim 6 , wherein, in the generation step,
 an initial value of a motion estimation process is set based on the relationship, and   the prediction image is generated by performing the motion estimation process using the initial value.   
     
     
         14 . An image decoding method in which the prediction image generation method according to  claim 5  is performed, the image decoding method comprising:
 an image decoding step of decoding the current block using the prediction image. 
 
     
     
         15 . The image decoding method according to  claim 14 , further comprising:
 a feature point information decoding step of decoding feature point information identifying the second feature point in a plurality of third feature points corresponding to the current block,   wherein the plurality of third feature points are extracted in the extraction step, and   the second feature point is selected from the plurality of third feature points using the feature point information in the search step.   
     
     
         16 . The image decoding method according to  claim 15 , wherein the feature point information indicates a coordinate of the second feature point. 
     
     
         17 . The image decoding method according to  claim 15 , wherein the feature point information indicates a rotation quantity or a scale value which is possessed by the second feature point. 
     
     
         18 . The image decoding method according to  claim 17 , further comprising:
 a corresponding point information decoding step of decoding corresponding point information identifying the corresponding point in the plurality of first feature points,   wherein the corresponding point is searched from the plurality of first feature points using the corresponding point information in the search step.   
     
     
         19 . The image decoding method according to  claim 18 , wherein, in the feature point information decoding step,
 indexes are allocated to the plurality of first feature points in a predetermined sequence, and   the corresponding point information indicates the index allocated to the corresponding point.   
     
     
         20 . A prediction image generation apparatus that generates a prediction image of a current block, the prediction image generation apparatus comprising:
 an extraction unit that extracts a plurality of first feature points each of which has a local feature quantity, the plurality of first feature points being included in a reconstructed image;   a search unit that searches a corresponding point from the plurality of first feature points, the corresponding point having a local feature quantity similar to a local feature quantity of a second feature point corresponding to the current block; and   a generation unit that generates the prediction image from the reconstructed image based on a relationship between the corresponding point and the second feature point being expressed by information including a non-parallel translation component.