ABSTRACT
In the context of imaging systems, we may often meet some constraints that limit the way that we typically transmit, store, and visualize the image content. In the case of black-and-white printing systems; a color image has to be converted into a meaningful gray-scale image. Another example is the photorealistic rendering on black-and-white media [141]; this system may reproduce the output color image, but it needs to be converted to a grayscale image for visualization purposes. The major aim of this conversion is to convey the same information contained in the original color image. This means preserving the appearance of the original color image while maintaining salient information such as edges, details, contrast, etc. The color2gray conversion problem can be seen as the classical dimensionality reduction problem, where an n-dimensional space is converted into an mdimensional space, where n > m. Unfortunately, traditional techniques, e.g., using the luminance channel Y as, gray-scale image, are often failing in preserving the appearance of the original color image, i.e., important features are lost. We will formalize the problem of color2gray conversion as the typical dimensionality reduction problem. Then we will introduce some na¨ıve approaches, followed by more sophisticated techniques with the aim of preserving specific features of the original color image. Finally, the inverse problem, colorization, will be briefly introduced.
