INTRODUCTION

My research is in the field of Image Processing. Because Image Processing is a huge field, the research is focussed in the so called low level image processing stage, which purpose is to improve the quality of the image, the extraction of features, etc. for its posterior use in a high level image processing stage.

IMAGE OPERATOR FOR DIRECTIONAL EDGE DETECTION

Our work proposes a basic image operator that perform a directional edge detection. Under certain conditions, the operator can perform a basic segmentation of superimposed patterns. Also, the operator can be used either as a detector or as an eraser. The proposed operator, as other edge detector operators, is based on the use of convolution masks. The values of the convolution masks will depend on the orientation of the edges of the pattern we want to erase.

Objective: Obtain a method to extract useful information from black and white images that have superimposed patterns.

Proposed Operator: The conventional process of detecting edges in an image is generally accomplished by applying a convolution mask to the entire image. This research show an alternative way to generate convolution masks dynamically. Once the convolution masks have been generated, the operator will be able to detect/erase basic concentric shapes. The values of the convolution masks depend on the orientation of the edge we want to detect/erase. The explanation becomes easier with an example. For this example please refer to the following figure.

FIGURE No. 1

To achieve our goal we need to establish a strong condition: to define a reference point O inside the image. This point will help us to determine the orientation of the edge in a given pixel and it will be the center of any concentric pattern we want to erase. Once the reference point has been defined, any pattern can be expressed as a set of points referenced to the central point. Suppose that we want to erase the circle shown in Fig.1. For a pixel P a convolution mask capable of erasing edges with the same orientation as line L should be generated.

The Base Function: The problem of detecting/erasing an edge, with a given orientation at pixel P is solved straightforward. The solution is to convolve P with a mask that detects/erases edges with an orientation q where q=d/2. To solved this problem we use a base function (defined as the derivatives of Gaussian) to sample the values for theconvolution masks. This is shown in Fig.2.

FIGURE No.2

Sampling Process: Once we have generated the appropiate base function, the values for the convolution mask are sampled from this function. This process is shown in Fig.3.

FIGURE No.3

Results: Some of the results we have obtained are shown. For more information please refer to the Section Presentations.