Method for image reframing

1. A method for processing an image according to one reframing method to obtain a reframed image for which the format ratio is equal to a target format ratio, the method comprising the steps of:

associating a saliency value with each pixel of said image;

applying a face detection method and a text detection method of said image

calculating a confidence parameter from said saliency values and from outputs of said face and text detection methods;

comparing the confidence parameter to a first predetermined threshold value;

selecting said reframing method in a set of at least two reframing methods according to the result of said comparison, and

reframing said image using the selected reframing method;

wherein, when said confidence parameter is less than said predefined threshold value, checking whether at least one zone of text is detected in said image and said reframing method is selected according the result of said checking;

wherein, when no text zone is detected, the processing method comprises further checking whether at least one face is detected and said reframing method is selected according to the result of said further checking.

2. The method according to claim 1 , wherein when said confidence parameter is greater than said first predefined threshold value, the reframing method selected is a method for dynamic reframing without deformation capable of delimiting a part of said image using a reframing window for which the format ratio is equal to the target format ratio.

3. The method according to claim 1 , wherein when a text zone is detected, the selected reframing method is a fixed reframing method with deformation.

4. The method according to claim 1 , wherein when at least two faces are detected and when said at least two faces are not capable of being comprised inside a reframing window of format ratio equal to the target format ratio then the reframing method selected is a dynamic reframing method with deformation, otherwise the reframing method selected is a fixed reframing method without deformation.

5. The method according to claim 1 , wherein when a single face is detected, if the average saliency of said image from which is subtracted the average saliency of said single face is less than a second predefined threshold value or greater than a third predefined threshold value, said second and third threshold values being defined such that said third threshold value is greater than said second threshold value then the reframing method selected is a fixed reframing method without deformation,

otherwise

if the saliency in the image is spread then the reframing method selected is a dynamic reframing method with deformation,

otherwise the reframing method selected is a fixed reframing method without deformation.

6. The method according to claim 1 , wherein when no face is detected, if the average saliency of said image is less than a second predefined threshold value or greater than a third predefined threshold value, said second and third threshold values being defined such that said third threshold value is greater than said second threshold value then the reframing method selected is a fixed reframing method without deformation,

otherwise

if the two pixels of the image of maximum saliency value are not capable of being comprised in a reframing window of format ratio equal to the target format ratio then the reframing method selected is a dynamic reframing method with deformation,

otherwise the reframing method selected is a fixed reframing method without deformation.

7. A device for reframing an image according to one reframing method to obtain a reframed image for which the format ratio is equal to a target format ratio, the device comprising

an analysis module of said image configured to associate a saliency value with each pixel of said image and to apply a face detection method and a text detection method on said image;

a calculation module configured to calculate a confidence parameter from said saliency values and from outputs of said face and text detection methods;

a comparator configured to compare the confidence parameter to a first predefined threshold value,

a selection module for selecting, for said image, said reframing method in a set of at least two reframing methods according to the result of the comparison carried out by the comparator; and

a reframing module for reframing said image using the selected reframing method;

wherein when said confidence parameter is less than said predefined threshold value, said selection module checks whether at least one zone of text is detected in said image and selects said reframing method according to the result of said check; and

wherein when no text zone is detected, said selection module further checks whether at least one face is detected and selects said reframing method according to the result of said further check.