Sidescan sonar image super resolution using sparse representation learning

With the rapid development of autonomous underwater vehicles, image produced from sidescan sonars mounting on these vehicles has attracted more and more attention. Many general super resolution approaches can not get satisfying results when applied to sidescan sonar image for their low levels of contrast. Compressive sensing theory suggested that high resolution image can be recovered from low resolution image with sparse representation. After training two dictionaries for high resolution and low resolution image patches, the sparse representation of given low resolution image can be applied to the dictionary of high resolution patches to get super resolution result image. Sparse representation based image super resolution approach applied to side scan sonar image is introduced. And the influence of patch size, dictionary size and filter is discussed in detail.