Fast image analogy analysis method based on compute unified device architecture (CUDA)

Abstract

The invention provides a fast image analogy analysis method based on a compute unified device architecture (CUDA). The method comprises the following steps that: 1) a source image A, a filtered image A' and a Gaussian pyramid of a source image B are generated; 2) an L-th layer is synthesized at present, and the following processes are executed for the pyramid from a coarse layer to the fine layer: 2.1) carrying out super sampling on B'L+1 to obtain B'L; 2.2) adopting a global search strategy eglobal based on a graphics processing unit (GPU) for each pixel point q in the B'L; 2.3) for each pixel point q in the B'L, adopting a k-coherence search strategy based on the GPU for finding consistency search optimal solution qkcoh; 2.4) giving aglobal as the value to qBest if the results obtained through the global search are better, otherwise, giving qkcoh as the value to the qBest, and storing the corresponding coordinates of the qBest in the A'L into the B'L(q); and 2.5) cyclically executing the steps 2.2) to 2.4) for setting the times; and 3) the B' is subjected to synthesis amplification to obtain high-resolution EHB'. The method has the advantages that the synthesis speed is accelerated, and the synthesis quality is improved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image analog synthesis method. Background technique [0002] Texture synthesis can automatically generate visually similar and continuous large sample textures from small area texture samples. The method of texture synthesis based on samples is a new texture collage technology developed after texture mapping and procedural texture synthesis. It avoids the problems of seams, aliasing, and repeated parameter adjustments in previous methods. At present, the research on pattern-based texture synthesis can be roughly divided into two aspects. On the one hand, it is hoped that more control factors can be added to the synthesis to constrain the texture generation and produce more diverse textures. On the other hand, it is desirable to speed up compositing so that compositing tends to be real-time. [0003] The image analogy framework borrows the idea of ​​analogy and provides a ver...

AI Technical Summary

Problems solved by technology

[0007] In order to overcome the disadvantages of slow synthesis speed and poor synthesis quality of existing image analog synthesis methods, the present invention provides a CUDA-based fast image analog synthesis method that improves synthesis speed and synthesis quality

Images