| 摘要 |
An inputted original image 100 is divided to a second block image by a second block generating section 2, and a second feature quantity is extracted by a second feature quantity extracting section 6. The second block image and the second feature quantity are stored in an image memory section 3. A control section 15 reads out, from the image memory section 3, only a block image having the same feature quantity as the feature quantity of a first block image from a first feature quantity extracting section 5, and transform processing is carried out by an image transforming and generating section 7 so as to send a block image 107 after transform to an approximate block search section 4. The approximate block search section 4 takes matching between the first block image 101 and the block image 107 after transform, and searches for a block image which realizes the minimum error between these block images. The approximate block search section 4 then selects and sends the block image to a coding and multiplexing section 9. Thus, in iterated image transform coding and decoding, a high coding efficiency can be maintained to provide an image of higher quality while approximate block search is carried out at a high speed. <IMAGE> |
