A system extracts a set of size invariant, rotation invariant features from pixel data of a character and sends the set of features to a statistical decision tree to effectuate automatic recognition of a character. The set of extracted features includes a group of features that is extracted by dividing a circle bounding the character into eight rings and 24 slices and counting the number of character pixels contained in each ring and in each slice. Various Fourier transforms, autocorrelations, moment calculations, and sorting operations are performed on many of the aforementioned arrays to yield further size and orientation invariant features. The distance between the centroid of the largest hole of the character and the center of the character is computed. The ratios of the perimeters of first and second holes of the character to the outer perimeter are computed.
If a polygon having an even number of vertexes, which are 2n in number, is input, a middle point calculation unit 24 pairs the m-th and (m+n)-th vertexes together, and calculates the coordinates of the middle points on the line segments between the vertexes of the respective pairs, which are n in number. A centroid calculation unit 25 calculates the coordinates of the centroid of the middle points. A distance calculation unit 26 calculates the distances between the centroid and the middle points. If all the distances between the centroid and the middle points are within a specified allowable range, a symmetry decision unit 27 sets the centroid as the center of point symmetry. If the centroid is selected as the center of point symmetry, a symmetrizing unit 28 corrects part or all of the vertexes to symmetrize the polygon with respect to the centroid.
There is disclosed an image processing device capable of receiving an image file having a first texture representation in rotation variant format and converting the texture representation to a modified texture representation in a rotation invariant format. The image processing device comprises an image processor for analyzing rotation variant texture parameters in the first texture representation and converting them to 1) rotation invariant texture parameters disposed in circular rings about a selected pixel; or 2)rotation invariant texture parameters disposed along radial lines extending through the selected pixel. In the second embodiment, the image processor can use the angular separation of the radial lines to determine the angular rotation of a rotated version of the image file.
Apparatus and methods are provided for object recognition and compression. The apparatus (114) comprises an object processor (402) configured to receive the image (120) and synthesize a contour (404) of an object within the image (120) and a classification engine (406) configured to receive the contour (404) of the image (120) and recognize the object within the image as a member of a first object class if the object substantially meets first object criteria of the first object class that is at least partially related to the target-specific utility of the image. The apparatus (114) also comprises a multi-rate encoder (116) configured to compress a first region of the image (120) having said object recognized as said member of said first object class at a first coding rate, said first coding rate providing a first coding resolution of said first region that is greater than a second coding resolution provided by a second coding rate for the image.
The present invention pertains to a method for sorting. The method comprises the steps of forming a decision tree comprised of at least a first level having at least a first entry with either a first result or a second result corresponding to values as they fill a queue. Then, there is the step of reading the result in the first entry of the first level of the decision tree. Next, there is the step of choosing the value in the first entry of the first level from the queue corresponding to the result in the first entry. The present invention comprises a priority queue. The priority queue comprises an array in which elements are stored. Also, the priority queue comprises a controller which stores the elements in the array with no memory reads and a constant number of memory writes so a smallest element in the array can be identified.
A method of representing an object appearing in a still or video image, by processing signals corresponding to the image, comprises deriving the peak values in CSS space for the object outline and applying a non-linear transformation to said peak values to arrive at a representation of the outline.
