Triangular Feature Extraction Algorithm for Image Compression and Recognition

摘要

PURPOSE: An algorithm of extracting triangular characteristics for image compression and recognition is provided to divide an image as maintaining position information of the image, thereby reflecting a shape of a face and positions of the eyes, ears, and the mouth on the image compression and recognition. CONSTITUTION: In a preprocessing step, the size of an image is regulated by dividing the image into the same size of 8 triangles. The size of the image is re-regulated to satisfy two conditions. For the first condition, the number of pixels in width and in length is even-numbered. For the second condition, the number of pixels in length is more than the number of pixels in width by 2. If the first condition is satisfied, a basic condition of dividing the image into the same size of 4 squares is satisfied. If the second condition is satisfied, a condition of dividing the 4 squares into the same size of 2 triangles is established. An 8 times 10 image satisfying the preprocessing condition is divided into the same size of 8 triangles. A method of obtaining data from each triangle is to extract features without reading position information by a regular setting like arrow directions.