Erasure-based quick response (QR) decoding
Systems and techniques are described herein for quick response (QR) code decoding. For example, a computing device can apply a filter to a QR code to blur the QR code; determine one or more areas of the filtered QR code that are associated with an error based on a light intensity variance of the one or more areas being above or below a variance threshold; and apply erasure correction to the one or more areas.
1 . An apparatus for quick response (QR) code decoding, the apparatus comprising:
at least one memory; and
at least one processor coupled to the at least one memory and configured to:
apply a filter to a QR code to blur the QR code;
determine one or more areas of the filtered QR code that are associated with an error based on a light intensity variance of the one or more areas being above or below a variance threshold; and
apply erasure correction to the one or more areas.
2 . The apparatus of claim 1 , wherein the erasure correction includes erasure-based Reed-Solomon correction.
3 . The apparatus of claim 1 , wherein the filter is a Gaussian filter.
4 . The apparatus of claim 1 , wherein the QR code is a reconstruction of an occluded QR code, wherein the reconstruction is based on application of white areas or black areas over one or more occlusions of the occluded QR code.
5 . The apparatus of claim 1 , wherein the at least one processor is further configured to:
determine the one or more areas of the filtered QR code are associated with the error based on the light intensity variance of the one or more areas being above or below the variance threshold using a machine learning model.
6 . An apparatus for quick response (QR) code decoding, the apparatus comprising:
at least one memory; and
at least one processor coupled to the at least one memory and configured to:
process a QR code, using a machine learning model, to generate an error confidence map including an indication that one or more areas of the QR code are associated with an error;
parse the error confidence map to determine a position of the one or more areas of the QR code associated with the error; and
apply erasure correction at the position of the one or more areas associated with the error.
7 . The apparatus of claim 6 , wherein the erasure correction includes erasure-based Reed-Solomon correction.
8 . The apparatus of claim 6 , wherein the at least one processor is further configured to:
adjust a size of the QR code to a reference size, wherein the machine learning model generates the error confidence map based on the QR code with the adjusted size and the error confidence map is based on the reference size; and
adjust the error confidence map from the reference size to the size of the QR code.
9 . The apparatus of claim 6 , wherein the at least one processor is further configured to:
select the machine learning model based on a size of the QR code.
10 . The apparatus of claim 6 , wherein the machine learning model includes a convolutional neural network (CNN).
11 . The apparatus of claim 6 , wherein the indication that the one or more areas of the QR code are associated with the error comprises probabilities that the one or more areas of the QR code are associated with the error.
12 . The apparatus of claim 6 , wherein the QR code is a reconstruction of an occluded QR code, wherein the reconstruction is based on application of white areas or black areas over one or more occlusions of the occluded QR code.
13 . The apparatus of claim 6 , wherein the at least one processor is further configured to:
determine, using the machine learning model, wherein the one or more areas of the QR code are associated with the error based on a light intensity variance of the one or more areas being above or below a variance threshold.
14 . A method for quick response (QR) code decoding, the method comprising:
processing a QR code, using a machine learning model, to generate an error confidence map including an indication that one or more areas of the QR code are associated with an error;
parsing the error confidence map to determine a position of the one or more areas of the QR code associated with the error; and
applying erasure correction at the position of the one or more areas associated with the error.
15 . The method of claim 14 , wherein the erasure correction includes erasure-based Reed-Solomon correction.
16 . The method of claim 14 , comprising:
adjusting a size of the QR code to a reference size, wherein the machine learning model generates the error confidence map based on the QR code with the adjusted size and the error confidence map is based on the reference size; and
adjusting the error confidence map from the reference size to the size of the QR code.
17 . The method of claim 14 , comprising:
selecting the machine learning model based on a size of the QR code.
18 . The method of claim 14 , wherein the machine learning model includes a convolutional neural network (CNN).
19 . The method of claim 14 , wherein the indication that the one or more areas of the QR code are associated with the error comprises probabilities that the one or more areas of the QR code are associated with the error.
20 . The method of claim 14 , wherein the QR code is a reconstruction of an occluded QR code, wherein the reconstruction is based on application of white areas or black areas over one or more occlusions of the occluded QR code.