System and method for image compression
This application relates to a method and a system for compressing a captured image of an item such as a mailpiece or parcel. The system may include a memory configured to store images of a plurality of items captured while the items are being transported and a processor in data communication with the memory. The processor may be configured to receive or retrieve one or more of the captured images, perform a wavelet scattering transform on the one or more captured images, perform deep learning on the wavelet scattering transformed images to classify the wavelet scattering transformed images and compress the classified wavelet scattering transformed images. Various embodiments can significantly improve a compression efficiency, a communication efficiency of compressed data and save a memory space so that the functionality of computing devices is significantly improved.
1. A system for compressing a captured image of an item, the system comprising:
a reader configured to capture an image of an item having a label thereon, the captured image comprising a grayscale image;
a memory configured to store the captured image of the item; and
one or more processors in data communication with the memory and the reader, the one or more processors configured to:
receive the grayscale image;
sum grayscale values of the grayscale image to produce a summed grayscale image;
average grayscale values of the summed grayscale image to produce a mean grayscale image;
transform the mean grayscale image using a transformation protocol;
classify the transformed mean grayscale image using a machine learning model; and
compress the classified transformed mean grayscale image.
2. The system of claim 1 , wherein to transform the mean grayscale image the one or more processors are further configured to:
perform one or more of wavelet transforms or a series of wavelet transforms on the mean grayscale image to produce one or more transformed images;
perform a nonlinearity operation on the transformed images; and
perform an averaging operation on the transformed images to produce a mean value of the transformed images on which the nonlinearity operation has been performed.
3. The system of claim 1 , wherein the one or more processors are further configured to:
identify features of the captured image distinguishable from each other; and
classify, using the machine learning model, the transformed mean grayscale image into two or more transformed images based on the identified features.
4. The system of claim 1 , wherein the captured image comprises an image of the label provided on an exterior surface of the item, the label comprising at least one of:
a return address region;
a mailing address region;
a barcode; a postage region; or
a specialty item region.
5. The system of claim 1 , wherein the one or more processors are configured to transform the summed grayscale image using the transformation protocol.
6. The system of claim 1 , wherein, to compress the classified transformed mean grayscale image, the one or more processors are configured to:
quantize values representing the classified transformed mean grayscale image;
compare the quantized values to a threshold, and discard values falling outside the threshold; and
encode remaining non-discarded quantized values to remove redundant information.
7. The system of claim 6 , wherein, in encoding the remaining non-discarded quantized values, the one or more processors are configured to perform at least one of:
entropy encoding, run-length encoding, or Huffman coding.
8. The system of claim 1 , wherein the captured image comprises binary data, and wherein the one or more processors is configured transform the binary data using the transformation protocol.
9. The system of claim 1 , wherein the classified transformed mean grayscale image comprises a plurality of features distinguishable from each other, and wherein the one or more processors are configured to compress at least part of the features of the classified transformed mean grayscale image.
10. A system comprising:
a reader configured to capture an image of an item having a label thereon;
a memory configured to store the captured image of the item; and
one or more processors in data communication with the memory and the reader, the one or more processors configured to:
receive a captured image;
perform a transformation on the captured image;
determine whether a value associated with the transformed image corresponds with a value for a machine learning model; and
train the machine learning model using data associated with the transformed image to classify the transformed image.
11. The system of claim 10 , wherein the one or more processors are further configured to:
extract one or more features from the transformed image; and
identify the machine learning model based on the one or more features.
12. The system of claim 11 , wherein the one or more features comprise a specific shape of the captured image.
13. The system of claim 11 , wherein the one or more features comprise one or more edges of the item.
14. The system of claim 10 , wherein the one or more processors are further configured to generate the machine learning model.
15. The system of claim 10 , wherein the data associated with the transformed image comprises one or more wavelet coefficients and a desired output; and wherein to train the machine learning model the one or more processors are configured to modify the one or more wavelet coefficients until the machine learning model converges on the desired output.
16. The system of claim 10 , wherein the memory is configured to store the data associated with the transformed image; and wherein to train the machine learning model the one or more processors are configured to retrieve the data associated with the image from the memory.
17. A method for compressing a captured image of an item, the method comprising:
receiving a captured image of an item having a label thereon from a reader, the captured image comprising a grayscale image;
summing grayscale values of the grayscale image to produce a summed grayscale image;
averaging grayscale values of the summed grayscale image to produce a mean grayscale image;
transforming the mean grayscale image using a transformation protocol;
classifying the transformed mean grayscale image using a machine learning model; and
compressing the classified transformed mean grayscale image.
18. The method of claim 17 , further comprising:
identifying features of the captured image distinguishable from each other; and
classifying, using the machine learning model, the transformed mean grayscale image into two or more transformed images based on the identified features.