Image quality assessment
This disclosure concerns image quality assessment. In particular, there is described a computer implemented method, software, and computer for assessing the quality of an image. For example but not limited to, quality of the image of a face indicates the suitability of the image for use in face recognition. The invention comprises determining ( 112 ) a similarity of features of two or more sub-images of the image ( 608 ) to a model ( 412 ) of the object which is based on multiple training images ( 612 ) of multiple different objects of that type. The model ( 412 ) is comprised of sub-models ( 406 ) and each sub-model ( 406 ) corresponds to a sub-image of the image ( 608 ). Determining similarity is based on the similarity of features of each sub-image to features modelled by the corresponding sub-model. It is an advantage that no input parameters are required for quality assessment since the quality of the image can be determined from only the similarity between the image and the same, therefore single generic, model.
1. A computer-implemented method for assessing quality of an image of an object, the method comprising:
determining a similarity of features of two or more sub-images of the image to a model of the object which is based on multiple training images of multiple different objects of a same type, the model is comprised of sub-models each modelled from multiple sub-images from the multiple training images, where each sub-model corresponds to a sub-image of the image, wherein determining similarity is based on the similarity of features of each sub-image to features modelled by the corresponding sub-model; and
determining the quality of the image based on the determined similarity of the two or more sub-images.
2. The computer-implemented method according to claim 1 , wherein the method is used as a pre-processing step for the image, and further processing is based on the determined quality of the image.
3. The computer-implemented method according to claim 1 , wherein each training image has multiple preferred characteristics that determine quality and a preferred characteristic includes one or more of a predetermined shift in the object, a predetermined rotation, a predetermined scale, a predetermined resolution, a predetermined pose, a predetermined illumination, and where the object is a face, a predetermined expression.
4. The computer-implemented method according to claim 1 , wherein each sub-model models features by a mean vector and a covariance matrix which are based on features of a sub-image of each training image that the sub-model is based on.
5. The computer-implemented method according to claim 1 , wherein the features of a sub-image are based on substantially only lower frequency components of the sub-image and the model is based on only lower frequency components of training images.
6. The computer-implemented method according to claim 1 , wherein the features modelled by a sub-model is a mean value of the features of a sub-image of each training image that the sub-model is based on.
7. The computer-implemented method according to claim 1 , wherein the features modelled by a sub-model are based on a Gaussian probability density function characterised by the mean vector and a covariance matrix of the features of a sub-image of each training image that the sub-model is based on.
8. The computer-implemented method according to claim 1 , wherein determining a similarity comprises determining a probability for the features of the sub-image of the image based on the Gaussian probability density function.
9. The computer-implemented method according to claim 1 , wherein the sub-image of the image is aligned with the sub-images of each training images the corresponding sub-model is based on.
10. The computer-implemented method according to claim 1 , wherein the determining a similarity is performed independently for each sub-image.
11. The computer-implemented method according to claim 1 , wherein determining similarity comprising determining one similarity measure for each sub-image and determining quality of the image is based on a combination of the determined similarity of the two or more sub-images.
12. The computer-implemented method according to claim 11 , wherein a combination of the determined similarities of the two or more sub-images comprises determining a probability for each sub-image independently and determining a joint probability based on the probabilities for each sub-image of the image.
13. The computer-implemented method according to claim 1 , wherein the method further comprises providing as output an indication of the determined quality of the image.
14. The computer-implemented method according to claim 1 , wherein the method further comprises storing an indication of the determined quality of the image in non-volatile memory.
15. The computer-implemented method according to claim 1 , wherein the method further comprises the steps of:
repeating the method for multiple different images to determine the quality of each of the multiple images; and
performing further processing of the images based on the determined quality.
16. The computer-implemented method according to claim 1 , wherein the object is a face.
17. The computer-implemented method according to claim 16 , wherein the further processing is performing face recognition or face verification on a subset of images determined as having better quality.
18. A non-transitory computer readable medium comprising instructions that when executed by a computer causes the computer to perform the method according to claim 1 .
19. A computer to assess the quality of an image of an object, the computer comprising:
computer storage to store a model of the object which is based on multiple training images of multiple different objects of a same type, the model is comprised of sub-models each modelled from multiple sub-images from the multiple training images, where each sub-model corresponds to a sub-image of the image; and
a processor to determine a similarity of features of two or more sub-images of the image to the model, wherein determining similarity is based on the similarity of features of each sub-image to features modelled by the corresponding sub-model, and to determine the quality of the image based on the determined similarity.