Parallel Fractal Compression Method for Big Video Data
Algorithm 3
Parallel Video Sequence Compression Algorithm Based on Fractal
Input:video sequence
Output:fractal code of video sequence
Fragment: video fragments are allocated to multiple processors for compression, which are divided according to the spatial–temporal similarity of video sequence.
Classify:Combine frames contained in video fragment into a large image matrix F for overall compression;
Divide F into non-overlapping N × N range blocks;
A 2 N × 2 N window with a step size of is used to intercept domain block along F;
Domain blocks are contracted by mean of neighboring four pixels;
According to Algorithm 1, domain blocks are classified and the corresponding range block categories are obtained;
Assign all categories of range block and corresponding domain blocks to multiple compute nodes for matching search;
Repeat
Remove a range block from the c-th range block category and set an initial value Error;
Repeat
Remove a domain block from the c-th domain block category, perform eight isometric transformations, and calculate according to Eqs.2–3 and Eq.10;
If then
Replace error with ;
End
Until All domain blocks in the c-th domain block category are completely searched;
Store fractal code {} of range block ;
Until All range block categories have been matched with their corresponding domain block category;
Merge: Combine the results of multiple compute nodes to obtain fractal code of video fragment ;
Merge:Combine the results of multiple processors to obtain fractal code of whole video ;