Image filtering method for recognizing the desired objects

Аннотация

The article formulates the task of preliminary processing of scene images to identify specific objects - route numbers of vehicles on color images obtained by shooting on camera devices. As accessible high-speed methods of pre-processing the Otsu method - a method of reception of optimum approach of the half-tone (grey) image, and also a number of standard methods of local processing in a sliding window is considered. The authors propose two heuristic pre-processing methods: nonadaptive transformation of a color image into gray (sharpness predictor); grey image blur method (BNS method - blurring the neighbors' shades) used to select the contours on the image. The test images show the effect of applying different methods to highlight contours and localize the area of the vehicle route number location on the images. The methods are also compared on other data - color photographs and a three-color cartographic image 1100 data with a dimension of 32х32 pixels.