Color-to-grayscale Image Transformation Preserving the Gradient Structure

A new algorithm for transforming a color image into a grayscale one is proposed. The algorithm accurately preserves the contrast range at the all image boundaries. Another application of the algorithm is visualization of vector fields. INTRODUCTION The need for proceeding from a more informative multicomponent image to a less informative singlecomponent one arises in a number of situations. First, for many high-performance technologies, the size of the data structures processed by the system is critical. The transition from a color to a monochrome image allows reducing the required computer memory size by a factor of 3. Second, some relevant image processing systems were initially designed to process only monochrome images. The conservative way of the evolution of such systems implies organizing an input filter that transforms the raw data to the form being standard for the system used. Third, effective expertise often requires visualizing a multi-component vector field, which is not so easy to achieve using conventional methods. Obviously, there is no general solution of the problem formulated. The solution substantially depends on the specific information that must be preserved when reducing a multi-component image. For example, it is often important to preserve the edges between different objects, that is, to keep the objects distinguishable. We shall demonstrate below that the conventional method of image grayscaling does not preserve the object edges and shall propose an original algorithm that is free from this drawback. PROBLEM STATEMENT The problem of color-to-grayscale image conversion is often solved by color coordinates transformation. For example, it is possible to proceed to so-called «brightness»: ( ) ( )