System and a method for semantic level image retrieval

A system for retrieval of similar images related to query images, the system comprising:

• a memory storing program instructions;
• a processor configured to execute program instructions stored in the memory;
• and
• an image processing engine in communication with the processor and configured to:
• pre-process the query images for noise removal by selecting a filtering technique for each query image based on a noise variance estimation in each of the query images with respect to a pre-set noise variance threshold value;
• pre-classify the pre-processed query images for determining a class one image identifier;
• generate one or more image types from the pre-processed query images for determining a class two image identifier;
• extract features from the pre-classified query images based on the class one image identifier and from the generated images based on the class two image identifier;
• retrieve one or more images similar to the query images, wherein the retrieved similar images have features similar to the extracted features of the pre-classified query images and the generated images;
• rank the retrieved similar images for determining the most similar images with respect to the query images;
• and
• analyze similarity between the query images and the retrieved similar images, wherein based on the analysis the retrieved similar images are re-ranked.

A computer program product comprising:

• a non-transitory computer-readable medium having computer program code store thereon, the computer-readable program code comprising instructions that, when executed by a processor, caused the processor to:
• pre-process the query images for noise removal by selecting a filtering technique for each query image based on a noise variance estimation in the query images with respect to a pre-set noise variance threshold value;
• pre-classify the pre-processed query images for determining a class one image identifier;
• generate one or more image types from the pre-processed query images for determining a class two image identifier;
• extract features from the pre-classified query images based on the class one image identifier and from the generated images based on the class two image identifier;
• retrieve one or more images similar to the query images, wherein the retrieved similar images have features similar to the extracted features of the pre-classified query images and the generated images;
• rank the retrieved similar images for determining the most similar images with respect to the query images;
• and
• analyze similarity between the query images and the retrieved similar images, wherein based on the analysis the retrieved similar images are re-ranked.