Neural networks arbitration for automatic edge detection of DNA bands in low-contrast images

Abstract

Low-contrast images, such as DNA autoradiograph images, provide a challenge for edge detection techniques, where the detection of the DNA bands within the images and locating their position is vital. In addition, the speed of recognition, high computational cost, and real-time implementation are also problems that haunt image processing. Thus, new measures are required to solve these problems. This paper reports on a new approach to solving the aforementioned problems. The novel idea is based on combining neural network arbitration and scale space analysis to automatically select one optimum scale for the entire image at which scale space edge detection can be applied. This approach to edge detection is formalised in the automatic edge detection scheme (AEDS). The AEDS is implemented on a real-life application namely, the detection of bands within low-contrast DNA autoradiograph images. An accurate comparison is drawn between the AEDS and the grammar-based multiscale analysis technique (GBMAT).