In the disclosed multi-view image encoding/decoding method in which a frame tobe encoded/decoded is divided and encoding/decoding is done to each region,first, aprediction image is generated not only for the region to be processed, butalso for thealready encoded/decoded regions neighboring to the region to be processed. Theprediction image is generated using the same prediction method for both kindsofregions. Next, correction parameters for correcting illumination and colormismatchesare estimated from the prediction image and decoded image of the neighboringregions.At this time, the estimated correction parameters can be obtained even at thedecodingside, therefore, encoding them is unnecessary. Thus, by using the estimatedcorrectionparameters to correct the predicted image that was generated for the region tobeprocessed, a corrected predicted image that can be actually used is generated.
展开▼
