High-speed parallel implementation method and device for template matching based on normalized correlation coefficient

Abstract

The invention discloses a high-speed parallel implementation method and device for template matching based on normalized correlation coefficients. The steps of the method are to read real-time graph and template graph data into the corresponding internal RAM buffer block and real-time graph data buffer RAM, and at the same time Carry out the calculation of the sum of the gray values ​​of the template image and the sum of the squares of the gray values ​​of the template image and the calculation of the sum of the squares of the gray values ​​of the real-time image at the search position (0,0); then simultaneously calculate the first row and subsequent columns of the search position The search position of the first row of each column template image real-time image gray value product sum and normalized correlation coefficient; further simultaneously read a new row of real-time image data into the corresponding internal RAM buffer block and the corresponding location of the real-time image data buffer RAM, At the same time, calculate the value of the first column of the current row; calculate the normalized correlation coefficient of each subsequent row in turn. The device is composed of a high-speed correlation arithmetic unit, an external data result memory and a microprocessor.

Description

technical field [0001] The invention relates to the technical field of image template matching, in particular to a high-speed parallel implementation method and device for template matching based on normalized correlation coefficients. Background technique [0002] Template matching is mainly used to locate a target in an image. It is widely used in image processing fields such as image alignment, edge detection, binocular stereo vision, etc., and these image processing methods have been used in missile target recognition, satellite image monitoring, medical image fusion, It is widely used in military and civilian fields such as measurement based on binocular stereo vision. [0003] Template matching mainly locates objects with similar sizes and images to the template map by calculating the similarity measure between the known template map at the search location and the real-time map search area. Because the normalized cross-correlation coefficient is invariant to brightnes...

AI Technical Summary

Problems solved by technology

[0009] The patent application number is 200910069272.6 The high-speed image matching method and device proposes a method to realize the normalized cross-correlation coefficient. This method has a simple structure, but the specific implementation process requires too many switches. Input, which will increase the consumption of logic resources, increase the difficulty of wiring, lead to longer wiring paths, reduce the maximum possible core speed of the system, and increase power consumption; the output is only the intermediate accumulation result, which will lead to the need for more external The storage space may require DSP to closely cooperate with the operation, thus occupying a large amount of DSP running time, resulting in the complexity of DSP control and communication, and reducing the real-time performance of DSP operation

Images