Leticia Vieira Guimaraes

Gradient Pile Up Algorithm for Edge Enhancement and Detection

Edge detection plays a fundamental role on image processing. The detected edges describe an object contour that greatly improves the pattern recognition process. Many edge detectors have been proposed. Most of them apply smooth filters to minimize the noise and the image derivative or gradient to enhance the edges. However, smooth filters produce ramp edges with the same gradient magnitude as those produced by noise. This work presents an algorithm that enhances the gradient correspondent to ramp edges without amplifying the noisy ones. Moreover, an efficient method for edge detection without set a threshold value is proposed. The experimental results show that the proposed algorithm enhances the gradient of ramp edges, improving the gradient magnitude without shifting the edge location. Further, we are testing the implementation of the proposed algorithm in hardware for real time vision applications.

Automatic generation of neural networks for image processing

This paper presents a technique for automatic generation of image processing architectures based on artificial neural networks (NN) for real time vision applications in order to reduce the hardware design effort. The generated datapath can be reused with different functions. A high throughput is obtained with one output pixel being produced at each clock cycle for each input pixel, allowing VGA stream processing. NN used is MLP, trained by back-propagation. Function training is executed in a C++ software. Then VHDL code of the image processing IP core is automatically generated. Image processing systems using the generated IP cores were evaluated in FPGA, showing both good performance and suitability of the method

Representação e Classificação de Texturas da Íris Baseada na Transformada Ótima de Gabor

This paper proposes to investigate the application of the method developed by Manjunah and Ma [12], namely, the Optimal Gabor Wavelet Transform (OGWT), in the context of iris texture representation. The proposed method was tested in a widely known database of 1205 eye images [16]. In each one of these images, the iris region was segmented, and then represented in multiple scales using the OGWT; the íris texture patterns were represented by their statistics in the Wavelet domain, and compared using similarity metrics. The experimental results indicate that the proposed 1 Programa de Pós-Graduação em Engenharia Elétrica – PPGEE, Universidade Federal do Rio Grande do Sul – UFRGS, Av. Oswaldo Aranha, 103, Porto Alegre, RS, Brasil {fernanda.quartieri@ufrgs.br, schuck@ufrgs.br, jacobs@inf.ufrgs.br, letícia@eletro.ufrgs.br} Representação e Classificação de Texturas da Íris Baseada na Transformada Ótima de Gabor 106 RITA • Volume XV • Número 2 • 2008 method obtains a correct iris recognition rate of 94,68%, even considering out of focus images and iris occlusions; a correct iris recognition rate of 100% is obtained excluding problematic images. The proposed method is flexible, and allows to fine tune the iris recognition criterion according to the accuracy level required by the application.

Gradient pile up for edge detection on hardware

Edge detection plays a fundamental role in image processing. Many edge detectors have been proposed. Most of them are based on the step edge model and applying smoothing filters to minimize the noise and the image derivative or gradient to enhance the edges. However, image sensors have a limited spatial bandwidth producing ramp edges with the same gradient magnitude as those produced by noise. This work presents a hardware implementation of the proposed edge detection algorithm. Our approach enhances the gradient correspondent to ramp edges without amplifying the noisy ones, for real time computer vision applications. The experimental results show that the proposed hardware implementation enhances the gradient of ramp edges, improving the gradient magnitude without shifting the edge location. Moreover, the execution time is appropriate for real time applications.