A method for face alignment operates on a face image and a set of initial landmark locations by first aligning globally the initial locations to a set of landmark locations of a face with a prototype shape to obtain global alignment parameters, and then warping the initial locations and the image from a coordinate frame of the image to a coordinate frame of the prototype shape according to the global alignment parameters to obtain warped landmark locations and a warped face image. Features are extracted from the warped face image at the warped landmark locations, and a regression function is applied to the features to obtain updated landmark locations in the coordinate frame of the prototype shape. Finally, the updated landmark locations in the coordinate frame of the prototype shape are warped to the coordinate frame of the image, to obtain updated landmark locations.

Asymmetries are detected in one or more images by partitioning each image to create a set of patches. Salient patches are identified, and an independent displacement for each patch is identified. The techniques used to identify the salient patches and the displacement for each patch are combined in a function to generate a score for each patch. The scores can be used to identify possible asymmetries.

In implementations of the subject matter as described herein, there is provided a method for forgery detection of a face image. Subsequent to inputting a face image, it is detected whether a blending boundary due to the blend of different images exists in the face image, and then a corresponding grayscale image is generated based on a result of the detection, where the generated grayscale image can reveal whether the input face image is formed by blending different images. If a visible boundary corresponding to the blending boundary exists in the generated grayscale image, it indicates that the face image is a forged image; on the contrary, if the visible boundary does not exist in the generated grayscale image, it indicates that the face image is a real image.

A method comprises capturing, with a vehicle vision sensor, a color image of a landing site including landing surface markers; converting the color image to a gray scale image; and performing multi-scale-binarization to detect multiple edges of the gray scale image and produce binary images. The method determines contours of edges of the binary images having closed shapes, detects closed shapes of contours of edges having four corners, and verifies whether four-sided candidate contours are valid as potential landing surface markers. If more than one contour is associated with a valid ID within a surface marker library, then the contour within the smallest window size is selected. If multiple contours with the same window size can be associated with a valid ID, then a mean of corresponding corners of multiple contours is computed. The method then performs corner refinement of valid four-sided candidate contours identified as potential landing surface markers.

An apparatus for detecting a mark by using image matching includes a damaged content obtaining unit configured to obtain damaged content generated based on original content including a plurality of cuts; a pre-processed content generating unit configured to generate pre-processed content by merging one or more images included in the damaged content or removing a partial region of one or more cuts included in the damaged content; a matchable content generating unit configured to generate matchable content, in which sizes or locations of one or more cuts included in the pre-processed content are adjusted by comparing one or more cuts included in the original image and the one or more cuts included in the pre-processed content; a matching unit configured to compare the original content with the matchable content, for each cut.

A method for determining an optimized weighting of an encoder and decoder network; the method comprising: for each of a plurality of test weightings, performing the following steps with the encoder and decoder operating using the test weighting: (a) encoding, using the encoder, a reference image and a distorted image into a latent space to form an encoding; (b) decoding the encoding, using the decoder, to form a distortion map indicative of a difference between the reference image and a distorted image; (c) spatially transforming the distorted image by the distortion map to obtain an aligned image; (d) comparing the aligned image to the reference image to obtain a similarity metric; and (e) determining a loss function which is at least partially defined by the similarity metric; wherein the optimized weighting is determined to be the test weighting which has an optimized loss function.

This invention provides a system and method for determining the pose of shapes that are known to a vision system that undergo both affine transformation and deformation. The object image with fiducial is acquired. The fiducial has affine parameters, including degrees of freedom (DOFs), search ranges and search step sizes, and control points with associated DOFs and step sizes. Each 2D affine parameter's search range and the distortion control points' DOFs are sampled and all combinations are obtained. The coarsely specified fiducial is transformed for each combination and a match metric is computed for the transformed fiducial, generating a score surface. Peaks are computed on this surface, as potential candidates, which are refined until a match metric is maximized. The refined representation exceeding a predetermined score is returned as potential shapes in the scene. Alternately the candidate with the best score can be used as a training fiducial.

A method and a system for rectifying a distorted fingerprint are provided. The method includes following steps. A feature of a distorted fingerprint is extracted, a reference distorted fingerprint whose feature is matched with the feature of the distorted fingerprint is searched for in a reference distorted fingerprint database, a dense distortion field of the reference distorted fingerprint is obtained in the reference distorted fingerprint database and the distorted fingerprint is rectified to a normal one according to the dense distortion field of the reference distorted fingerprint.

Example image processing methods, apparatus/systems and articles of manufacture are disclosed herein. An example apparatus includes an image recognition application to identify matches between stored patterns and objects detected in a shelf image, where the shelf image has a shelf image scale estimate. The example apparatus further includes a scale corrector to calculate deviation values between sizes of (A) a first set of the objects detected in the shelf image and (B) a first set of the stored patterns matched with the first set of the objects and reduce an error of the shelf image scale estimate by calculating a scale correction value for the shelf image scale estimate based on the deviation values.

In implementations of the subject matter as described herein, there is provided a method for forgery detection of a face image. Subsequent to inputting a face image, it is detected whether a blending boundary due to the blend of different images exists in the face image, and then a corresponding grayscale image is generated based on a result of the detection, where the generated grayscale image can reveal whether the input face image is formed by blending different images. If a visible boundary corresponding to the blending boundary exists in the generated grayscale image, it indicates that the face image is a forged image; on the contrary, if the visible boundary does not exist in the generated grayscale image, it indicates that the face image is a real image.