Documentation

Array containing a grayscale image (color images are not supported)

FAST threshold for which a pixel of the circle around the central pixel is considered to be greater or smaller

Length of arc (or sequential segment) to be tested, must be within range [9-16]

Performs non-maximal suppression if true

Maximum ratio of features to detect, the maximum number of features is calculated by feature_ratio * in.elements(). The maximum number of features is not based on the score, instead, features detected after the limit is reached are discarded

Is the length of the edges in the image to be discarded by FAST (minimum is 3, as the radius of the circle)

Struct containing arrays for x and y coordinates and score, while array orientation is set to 0 as FAST does not compute orientation, and size is set to 1 as FAST does not compute multiple scales

array containing a grayscale image (color images are not supported)

maximum number of corners to keep, only retains those with highest Harris responses

minimum response in order for a corner to be retained, only used if max_corners = 0

the standard deviation of a circular window (its dimensions will be calculated according to the standard deviation), the covariation matrix will be calculated to a circular neighborhood of this standard deviation (only used when block_size == 0, must be >= 0.5f and <= 5.0f)

square window size, the covariation matrix will be calculated to a square neighborhood of this size (must be >= 3 and <= 31)

struct containing arrays for x and y coordinates and score (Harris response), while arrays orientation and size are set to 0 and 1, respectively, because Harris does not compute that information

Array containing a grayscale image (color images are not supported)

FAST threshold for which a pixel of the circle around the central pixel is considered to be brighter or darker

maximum number of features to hold (will only keep the max_feat features with higher Harris responses)

factor to downsample the input image, meaning that each level will hold prior level dimensions divided by scl_fctr

number of levels to be computed for the image pyramid

blur image with a Gaussian filter with sigma=2 before computing descriptors to increase robustness against noise if true

Features struct composed of arrays for x and y coordinates, score, orientation and size of selected features

Array containing a grayscale image (color images are not supported)

number of layers per octave, the number of octaves is computed automatically according to the input image dimensions, the original SIFT paper suggests 3

threshold used to filter out features that have low contrast, the original SIFT paper suggests 0.04

threshold used to filter out features that are too edge-like, the original SIFT paper suggests 10.0

the sigma value used to filter the input image at the first octave, the original SIFT paper suggests 1.6

if true, the input image dimensions will be doubled and the doubled image will be used for the first octave

the inverse of the difference between the minimum and maximum grayscale intensity value, e.g.: if the ranges are 0-256, the proper intensity_scale value is 1256, if the ranges are 0-1, the proper intensity-scale value is 11

maximum ratio of features to detect, the maximum number of features is calculated by feature_ratio * in.elements(). The maximum number of features is not based on the score, instead, features detected after the limit is reached are discarded

Features object composed of arrays for x and y coordinates, score, orientation and size of selected features Nx128 array containing extracted descriptors, where N is the number of features found by SIFT

Array containing a grayscale image (color images are not supported)

number of layers per octave, the number of octaves is computed automatically according to the input image dimensions, the original SIFT paper suggests 3

threshold used to filter out features that have low contrast, the original SIFT paper suggests 0.04

threshold used to filter out features that are too edge-like, the original SIFT paper suggests 10.0

the sigma value used to filter the input image at the first octave, the original SIFT paper suggests 1.6

if true, the input image dimensions will be doubled and the doubled image will be used for the first octave

the inverse of the difference between the minimum and maximum grayscale intensity value, e.g.: if the ranges are 0-256, the proper intensity_scale value is 1256, if the ranges are 0-1, the proper intensity-scale value is 11

maximum ratio of features to detect, the maximum number of features is calculated by feature_ratio * in.elements(). The maximum number of features is not based on the score, instead, features detected after the limit is reached are discarded

Features object composed of arrays for x and y coordinates, score, orientation and size of selected features ^ Nx272 array containing extracted GLOH descriptors, where N is the number of features found by SIFT

is the Array containing the data to be queried

is the Array containing the data used as training data

indicates the dimension to analyze for distance (the dimension indicated here must be of equal length for both query and train arrays)

is the number of smallest distances to return (currently, only 1 is supported)

is an array of MxN size, where M is equal to the number of query features and N is equal to n_dist. The value at position IxJ indicates the index of the Jth smallest distance to the Ith query value in the train data array. the index of the Ith smallest distance of the Mth query. is an array of MxN size, where M is equal to the number of query features and N is equal to n_dist. The value at position IxJ indicates the Hamming distance of the Jth smallest distance to the Ith query value in the train data array.

is the array containing the data to be queried

is the array containing the data used as training data

indicates the dimension to analyze for distance (the dimension indicated here must be of equal length for both query and train arrays)

is the number of smallest distances to return (currently, only values <= 256 are supported)

is the distance computation type. Currently AF_SAD (sum of absolute differences), AF_SSD (sum of squared differences), and AF_SHD (hamming distances) are supported.

is an array of MxN size, where M is equal to the number of query features and N is equal to n_dist. The value at position IxJ indicates the index of the Jth smallest distance to the Ith query value in the train data array. the index of the Ith smallest distance of the Mth query. is an array of MxN size, where M is equal to the number of query features and N is equal to n_dist. The value at position IxJ indicates the distance of the Jth smallest distance to the Ith query value in the train data array based on the dist_type chosen.

is an Array with image data

is the template we are looking for in the image

is metric that should be used to calculate the disparity between window in the image and the template image. It can be one of the values defined by the enum af_match_type

will have disparity values for the window starting at corresponding pixel position

is input grayscale/intensity image

nucleus radius for each pixel neighborhood

intensity difference threshold a.k.a t from equations in description

geometric threshold

is maximum number of features that will be returned by the function

indicates how many pixels width area should be skipped for corner detection

is input image

is the radius of first gaussian kernel

is the radius of second gaussian kernel

is difference of smoothed inputs

x coordinates of the source points.

y coordinates of the source points.

x coordinates of the destination points.

y coordinates of the destination points.

htype, can be AF_HOMOGRAPHY_RANSAC, for which a RANdom SAmple Consensus will be used to evaluate the homography quality (e.g., number of inliers), or AF_HOMOGRAPHY_LMEDS, which will use Least Median of Squares method to evaluate homography quality.

If htype is AF_HOMOGRAPHY_RANSAC, this parameter will five the maximum L2-distance for a point to be considered an inlier.

maximum number of iterations when htype is AF_HOMOGRAPHY_RANSAC and backend is CPU, if backend is CUDA or OpenCL, iterations is the total number of iterations, an iteration is a selection of 4 random points for which the homography is estimated and evaluated for number of inliers.

is a 3x3 array containing the estimated homography. is the number of inliers that the homography was estimated to comprise, in the case that htype is AF_HOMOGRAPHY_RANSAC, a higher inlier_thr value will increase the estimated inliers. Note that if the number of inliers is too low, it is likely that a bad homography will be returned.