Systems and methods for barcode annotations for digital images
A content-based image retrieval (CBIR) system and method is presented herein. The CBIR system generates a relatively short vector or array of data, referred to as a barcode, from an input image. The short vector or array data can be used to represent the content of the image for image retrieval purposes. The system obtains the image and applies a transform to the image to generate a plurality of image transform values. The system thresholds the plurality of image transform values to obtain compact image transform values. The system generates a barcode in accordance with the compact image transform values and representative of the image. The system may then transmit the barcode to a database for storage or draw the barcode on a display. The system may also compare barcodes to find and retrieve similar images associated with similar barcodes.
1. A method of identifying a similar image for a query image, the method comprising:
storing, in a database, a plurality of comparison barcodes indexed with respective comparison images; and
operating a processor to:
obtain the query image;
select one or more regions of interest (ROI) from a set of regions of interest identified by a user or a set of regions of interest automatically identified by the processor for selection;
generate one or more query barcodes for the one or more regions of interest to represent the query image, wherein generating the one or more query barcodes comprises:
applying a transform to at least a portion of the query image to extract a subset of image data and to generate a plurality of image transform values based on the extracted subset of image data;
thresholding the plurality of image transform values to generate a plurality of compact image transform values, the plurality of compact image transform values corresponding to the plurality of image transform values having a reduced range of values; and
generating the one or more query barcodes based on the plurality of compact image transform values to represent the query image;
retrieve, from the database, one or more comparison barcodes from the plurality of comparison barcodes;
calculate a signal distance between each comparison barcode and the query barcode to determine a degree of similarity between that comparison barcode and the one or more query barcodes; and
identify the similar image for the query image by identifying the comparison image indexed with the comparison barcode that has the shortest signal distance to the one or more query barcodes.
2. The method of claim 1 , wherein the transform is a Radon transform.
3. The method of claim 2 comprising operating the processor to:
select a number of projections for the Radon transform;
apply noise reduction to the query image;
under-sample the image; and
normalize pixel intensities of the query image.
4. The method of claim 3 comprising operating the processor to threshold the plurality of image transform values via at least one of: local thresholding, global thresholding, incremental thresholding, and min-max thresholding.
5. The method of claim 1 comprising operating the processor to:
retrieve the comparison image associated with the comparison barcode that has the shortest signal distance to the query barcode; and
output that comparison image.
6. The method of claim 1 , wherein the signal distance is a Hamming distance.
7. The method of claim 1 comprising operating the processor to compress the plurality of image transform values using an artificial neural network.
8. The method of claim 1 comprising operating the processor to generate at least one query barcode to represent the query image by generating one or more arrays of data based on the query image.
9. A method of identifying a similar image for a query image, the method comprising:
storing, in a database, a plurality of comparison barcodes indexed with respective comparison images; and
operating a processor to:
obtain the query image;
select one or more regions of interest (ROI) from a set of regions of interest identified by a user or a set of regions of interest automatically identified by the processor for selection;
generate a set of query barcodes for the one or more regions of interest to represent the query image, wherein generating the set of query barcodes comprises:
selecting a number of projections for a Radon transform;
applying the Radon transform to at least a portion of the query image to extract a subset of image features and to generate a plurality of Radon projection functions based on the extracted subset of image features;
thresholding the plurality of Radon projection functions to generate a plurality of compact Radon projection functions, the plurality of compact Radon projection functions corresponding to the plurality of Radon projection functions having a reduced range of values; and
generating the set of query barcodes based on the plurality of compact Radon projection functions, each query barcode being generated based on a compact Radon projection function of the plurality of compact Radon projection functions;
retrieve, from the database, one or more comparison barcodes from the plurality of comparison barcodes;
calculate a signal distance between each comparison barcode and each query barcode to determine a similarity between the respective comparison barcode and the respective query barcode;
for each comparison barcode, sum all signal distances calculated from the comparison with the set of query barcodes to generate a total signal distance value; and
identify the similar image for the query image by identifying the comparison image indexed with the comparison barcode that has the shortest total signal distance.
10. The method of claim 9 comprising operating the processor to threshold the plurality of Radon projection functions via at least one of: local thresholding, global thresholding, incremental thresholding, and min-max thresholding.
11. The method of claim 9 comprising operating the processor to:
retrieve the comparison image associated with the comparison barcode that has the shortest total signal distance to the set of query barcodes; and
output the comparison image.
12. The method of claim 9 , wherein the signal distance is a Hamming distance.
13. The method of claim 9 comprising operating the processor to compress the plurality of Radon projection functions using an artificial neural network.
14. The method of claim 9 comprising operating the processor to generate at least one query barcode to represent the query image by generating one or more arrays of data based on the query image.
15. A method of identifying at least one similar image for a query image, the method comprising:
storing, in a database, a plurality of comparison image representations indexed with respective comparison images; and
operating a processor to:
obtain the query image;
select one or more regions of interest (ROI) from a set of regions of interest identified by a user or a set of regions of interest automatically identified by the processor for selection;
generate one or more query image representations for the one or more regions of interest to represent the query image, wherein generating the one or more query image representations comprises:
applying a transform to at least a portion of the query image to extract a subset of image data and to generate a plurality of image transform values based on the extracted subset of image features;
thresholding the plurality of image transform values to generate a plurality of compact image transform values, the plurality of compact image transform values corresponding to the plurality of image transform values having a reduced range of values;
generating the one or more query image representations based on the plurality of compact image transform values to represent the query image;
retrieve, from the database, one or more comparison image representations from the plurality of comparison image representations;
calculate a signal distance between each comparison image representation and the one or more query image representations to determine a degree of similarity between that comparison image representation and the one or more query image representations; and
identify the at least one similar image for the query image by identifying the comparison image indexed with the comparison image representation that has the shortest signal distance to the query image representation.
16. The method of claim 15 , wherein the transform is a Radon transform.
17. The method of claim 16 comprising operating the processor to:
select a number of projections for the Radon transform;
apply noise reduction to the query image;
under-sample the image; and
normalize pixel intensities of the query image.
18. The method of claim 17 comprising operating the processor to threshold the plurality of image transform values via at least one of: local thresholding, global thresholding, incremental thresholding, and min-max thresholding.
19. The method of claim 15 , wherein the signal distance is a Hamming distance.
20. The method of claim 15 comprising operating the processor to compress the plurality of image transform values using an artificial neural network.