Method for determining a colour value of an object in an image

Patent metadata

Field	Value
Publication number	US-10380415-B2
Application number	US-201715426710-A
Country	US
Kind code	B2
Filing date	Feb 7, 2017
Priority date	Feb 12, 2016
Publication date	Aug 13, 2019
Grant date	Aug 13, 2019

Abstract

A method for determining an apparent color value of pixels corresponding to a subpart of an object in the acquired image, including: estimating a geometric distortion model of the object in the acquired image compared with a reference model of the object; determining a position in the acquired image of the subpart and determining a zone of the acquired image, referred to as the reference zone, situated in a predetermined vicinity of the subpart, each pixel of the reference zone resulting from a projection in the acquired image of a zone of the reference model, referred to as a blank zone, the blank zone being associated with a predetermined color value in the reference model, the projection using the geometric distortion model; and determining an apparent color value of the pixels of the subpart, from a color value of at least one pixel of the reference zone.

First claim

The invention claimed is:

1. A method for determining an apparent colour value of pixels corresponding to a subpart of an object represented in an image, referred to as the acquired image, wherein the method comprises:
estimating, by a processor, a geometric distortion model of said object in the acquired image compared with a reference model of the object, the object in the acquired image being a distorted version of the object in the reference model, the geometric distortion model allowing passing from the object in the acquired image to the object in the reference model and conversely;
from the geometric distortion model and the reference model of the object, determining, by the processor, a position of the subpart in the acquired image, called acquired subpart, from a position of the same subpart in the reference model, called model subpart, and determining a zone of the acquired image, referred to as the reference zone, situated in a predetermined vicinity of the acquired subpart, from the same zone in the reference model, referred to as a blank zone, the blank zone being associated with a predetermined colour value in the reference model of the object; and
determining, by the processor, an apparent colour value of the pixels of the acquired subpart, from a colour value of at least one pixel of the reference zone.

2. A method for determining that a mark has been applied to a subpart of a document, said document comprising at least one subpart on which a mark can be applied, an image representing the document, referred to as the acquired image, having been acquired by an image acquisition device, wherein the method comprises:
for each subpart of the document on which a mark can be applied, determining an apparent colour value of the pixels of the acquired image corresponding to the subpart by applying the method according to claim 1 to the subpart, the object being the document; and
deciding, for each subpart of the document on which a mark can be applied, that a mark has been applied on said subpart when a proportion of the pixels corresponding to the subpart in the acquired image having a colour value, referred to as the pixel colour value, close to the apparent colour value, is below a first predetermined threshold, a pixel colour value being close to the apparent colour value when a difference between said pixel colour value and the apparent colour value is less than a predetermined difference.

3. The method according to claim 2 , wherein the method further comprises applying a correction to the pixels of the acquired image corresponding to each subpart of the document on which a mark can be applied according to the apparent colour value obtained for said subpart, each pixel on which a correction is applied becoming a corrected pixel, each pixel taken into account for determining whether the proportion of pixels corresponding to the subpart having a pixel colour value close to the apparent colour value is below the first predetermined threshold being a corrected pixel.

4. The method according to claim 3 , wherein the correction applied to the pixels of the acquired image corresponding to a subpart of the document on which a mark can be applied comprises applying a different normalisation factor to each component of each of said pixels, each weighting factor being calculated from the apparent colour value associated with said subpart.

5. The method according to claim 2 , wherein the method further comprises filtering among the apparent colour values determined for each subpart of the document on which a mark can be applied, aberrant colour values, an apparent colour value of a first subpart of a document being considered to be aberrant when a difference between the apparent colour value and a colour value, referred to as the vicinity colour value, representing apparent colour values of second subparts, adjacent to the first subpart, is above a second predetermined threshold, a second subpart being adjacent to a first subpart when a distance between the first subpart and the second subpart is less than a predetermined distance, the apparent colour value associated with a subpart being replaced by the vicinity colour value when the apparent value is an aberrant colour value.

6. The method according to claim 2 , wherein each subpart of the document is a box delimited by a contour and in that the method further comprises, for each box on the document:
determining, from the distortion model and the reference model of the document, a position of each pixel corresponding to the contour of the box;
determining a colour value, referred to as the contour colour value, representing colour values of pixels of said contour; and
correcting the colour value of each pixel of said contour and/or of each pixel situated inside said contour using the contour colour value and a contour pixel colour value given by the reference model of the document.

7. The method according to claim 6 , wherein, when a pixel of said contour or a pixel situated inside said contour is corrected, the pixel is transformed so as to represent it in a colorimetric space in which each pixel is associated with components comprising a hue component, the correction consisting of modifying the hue component using a hue component issuing from the contour colour value and a hue component issuing from the contour pixel colour value given by the reference model of the document.

8. A method for correcting an image, referred to as the acquired image, of a document, said document comprising at least one subpart on which a mark can be applied, wherein the method:
for each subpart of the document on which a mark can be applied, determining an apparent colour value of the pixels of the acquired image corresponding to the subpart by applying the method according to claim 1 to the subpart, the object being the document; and
applying a correction to the pixels of the acquired image corresponding to each subpart of the document according to the apparent colour value obtained for said subpart.

9. The method according to claim 8 , wherein, the correction applied to the pixels of the acquired image corresponding to a subpart of the document comprises applying a different normalisation factor to each component of each of said pixels, each weighting factor being calculated from the apparent colour value associated with said subpart.

10. The method according to claim 8 , wherein the method further comprises filtering among the apparent colour values determined for each subpart of the document on which a mark can be applied, aberrant colour values, an apparent colour value of a first subpart being considered to be aberrant when a difference between the apparent colour value and a colour value, referred to as the vicinity colour value, representing apparent colour values of second subparts adjacent to the first subpart, is above a predetermined threshold, a second subpart being adjacent to a first subpart when a distance between the first subpart and the second subpart is less than a predetermined distance, the apparent colour value associated with a subpart being replaced by the vicinity colour value when the apparent colour value is an aberrant colour value.

11. The method according to claim 8 , wherein each subpart of the document is a box delimited by a contour and the in that method further comprises, for each box on the document:
determining, from the distortion model and the reference model of the document, a position of each pixel corresponding to the contour of the box;
determining a colour value, referred as the contour colour value, representing colour values of pixels of said contour; and
correcting the colour value of each pixel of said contour and/o

Assignees

• Safran Identity & Security

Inventors

• Rostaing Laurent
• Rouh Alain
• Beaudet Jean

Classifications

• G06V10/56

  relating to colour · CPC title

• G06V10/247

  by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids · CPC title

• G06V10/30

  Noise filtering · CPC title

• G06T2207/30176

  Document · CPC title

• G06T7/75

  involving models · CPC title