Method and system for automatic selection of one or more image processing algorithm

Patent metadata

Field	Value
Publication number	US-9275307-B2
Application number	US-201414286303-A
Country	US
Kind code	B2
Filing date	May 23, 2014
Priority date	May 24, 2013
Publication date	Mar 1, 2016
Grant date	Mar 1, 2016

Abstract

Disclosed is a method and system for automatic algorithm selection for image processing. The invention discloses the method and system for automatically selecting the correct algorithm(s) for a varying requirement of the image for processing. The selection of algorithm is completely automatic and guided by a plurality of machine learning approaches. The system here is configured to pre-process plurality of images for creating a training data. Next, the test image is extracted, pre-processed and matched for assessing the best possible match of algorithm for processing.

First claim

What is claimed is:

1. A computer implemented method for automatically selecting one or more algorithms in an image processing environment, the method comprising:
receiving at least one image;
segmenting each image received into a plurality of image segments;
performing one or more pre-stored algorithms from a plurality of image processing algorithms upon each of the image segments to obtain a plurality of image processing algorithm outputs;
comparing each of the image processing algorithm outputs against a predetermined threshold image processing output score;
for each of the image processing algorithm outputs exceeding the predetermined threshold image processing output score, recording the image processing algorithm along with corresponding one or more image segments and associated feature vectors as a training pair;
receiving one or more incoming pre-processed test images; and
selecting one or more potentially matching image processing algorithms from the training pair for each incoming pre-processed test image, wherein the potentially matching image processing algorithm is selected by matching a test segment of the incoming test image and a feature vector associated with the test segment with the one or more segments recorded and feature vectors associated with the one or more segments recorded.

2. The computer implemented method of claim 1 , further comprising, to prepare the pre-processed test image:
receiving a test image;
segmenting the test image into a plurality of test image segments; and
extracting a plurality of features from the plurality of test image segments.

3. The computer implemented method of claim 1 , wherein the image segments include overlapping segments, non-overlapping segments, or a combination of thereof.

4. The computer implemented method of claim 1 , wherein the plurality of image processing algorithms include binarization algorithms, noise reduction algorithms, segmentation algorithms, character recognition algorithms, page layout detection, page layout segmentation, or a combination thereof.

5. The computer implemented method of claim 1 , further comprising obtaining the feature vectors by:
extracting image moment features for each of the image segments, wherein the image moment features comprise a mean, a variance, and a skew corresponding to the image segment; and
forming the feature vectors using HSV quantization.

6. The computer implemented method of claim 1 , wherein the predetermined threshold image processing output score is a groundtruth determined as a characteristic for each training pair.

7. An image processing system for automatically selecting one or more algorithms in an image processing environment, the system comprising:
a processor;
a memory coupled to the processor, wherein the processor is configured to executing a plurality of modules stored in the memory, and wherein the plurality of modules comprise:
a receiving module to receive at least one image;
an image processing module that:
segments each image received into a plurality of image segments; and
performs one or more pre-stored algorithms from a plurality of image processing algorithms upon each of the image segments in order to obtain a plurality of image processing algorithm outputs;

a comparing module that compares each of the image processing algorithm outputs against a predetermined threshold image processing output score;
a recording module that records the image processing algorithms, those exceeding the predetermined threshold image processing output score, along with corresponding one or more image segments and associated feature vectors as a training pair; and
a matching module that selects one or more potentially matching image processing algorithms from the training pair for an incoming pre-processed test image, wherein the potentially matching image processing algorithm is selected by matching a test segment of the incoming test image and a feature vector associated with the test segment with the one or more segments recorded and feature vectors associated with the one or more segments recorded.

8. image processing system of claim 7 , wherein the image processing module processes the at least one image by performing one or more image processing algorithms selected from binarization algorithms, noise reduction algorithms, segmentation algorithms, character recognition algorithms, page layout detection, page layout segmentation, or a combination thereof.

9. The image processing system of claim 7 , wherein the incoming pre-processed test image comprises a plurality of test image segments thereof and extracted features from each test image segment.

10. image processing system of claim 7 , wherein the recording module extracts image moment features for each of the image segments, wherein the image moment features comprise a mean, a variance, and a skew corresponding thereto; and forms the feature vectors using HSV quantization.

11. The image processing system of claim 7 , wherein the comparing module determines the predetermined threshold image processing output score as a groundtruth that is characteristic for each training pair.

12. A non-transitory computer readable medium embodying a program executable in a computing device for automatically selecting one or more algorithms in an image processing environment, the program comprising a set of instructions for:
receiving at least one image;
segmenting each image received into plurality of image segments;
performing one or more pre-stored algorithms from plurality of image processing algorithms upon each of the image segment in order to obtain plurality of image processing algorithm output;
comparing each of the image processing algorithm output against a predetermined threshold image processing output score;
for each of the image processing algorithm exceeding the predetermined threshold image processing output score, recording the image processing algorithm along with corresponding one or more image segments and associated feature vectors as a training pair; and
selecting one or more potentially matching image processing algorithm from the training pair for each incoming pre-processed test image, wherein the potentially matching image processing algorithm is selected by matching a test segment of the incoming test image and a feature vector associated with the test segment with the one or more segments recorded and feature vectors associated with the one or more segments recorded.

Assignees

• Tata Consultancy Services Ltd

Inventors

• Chattopadhyay Tanushyam
• Reddy Ramu Vempada
• Garain Utpal

Classifications

• G06V30/18105

  related to colour · CPC title

• G06V30/162 Primary

  Quantising the image signal · CPC title

• G06V30/10

  Character recognition · CPC title

• G06K9/38

  Physics · mapped topic

• G06K9/4652

  Physics · mapped topic