Detection and comparison of color edges via median based PCA

In this paper we present a novel image quality measure for color images. This measure is based on the correlation of information between color planes and principal component analysis conversion of a color image into grayscale. Furthermore, a scheme for boundary detection based on this principle is developed and presented. This innovative design utilizes partial derivatives of Boolean functions for edge detection and is then analyzed using our new measure. Results testing the scheme on a database of natural and synthetic images show the performance to be quantitatively competitive with existing measures.