Work

Non Linear Algorithm for Automatic Image Processing Applications in FPGAs

Year: 2011
Type: book-chapter
Abstract: Mathematical morphology supplies powerful tools for low-level image analysis. The design of morphological operators for a given application is not a trivial one. For some problems in low level image processing the best result is achieved applying to the image an ordered sequence of morphological operators, that can be done manually, but is not easy and not always leads to the best solution. Genetic programming (GP) is a branch of evolutionary computing, and it is consolidating as a promising method for applications of digital image processing. The main objective of genetic programming is to discover how computers can learn to solve problems without being programmed for that. In the search for a practical automatic solution for low level image processing using mathematical morphology and genetic programming we present in this chapter a Matlab algorithm used for this purpose. Two sample images feed the Matlab application, the first one the original image with all defects, the second one the goal image where the defects of the original image were corrected. If we want to find the mathematical morphology operators that implement a certain filter that removes specific noise from the image, we supply a noisy image and an image were the noise was removed. The second image can be obtained from the noisy image applying an image manipulation program. After the parameters are supplied to the Matlab algorithm, the developed program starts to search for the sequence of morphological operators that leads to the best solution. The program works iteratively, and at each iteration compares the result of the morphological operations applied to the image set with the previous ones. To quantify how good is the solution at each iteration the resulting image is compared with the reference image using the mean absolute error (MAE) of the pixels. The best solution of the process is the image from a certain set whose error is less than a reference error indicated to the function. Using this methodology it was possible to solve a number of low level image processing problems, including edge detection, noise removal, separation of text from figures, with an error less than 0.5%, most of the time. Examples are presented along the text to clarify the use of the proposed algorithm. In addition, the sequence of operators obtained by the Matlab procedure was used to reconfigure an hardware architecture implemented in FPGAs to process images with the generated instructions in real time.
Language: English
Open Access status: hybrid
OpenAlex ID: W1532524255
FWCI:
Cites: 16
Cited by:
Related to: 10