Perceptual Quality of Color Images of Natural Scenes Transformed in CIELUV Color Space

Transformations of digitized color images in perceptually-uniform CIELUV color space and their perceptual relevance were investigated. Chroma variation was chosen as the first step of a series of investigations into possible transformations (including lightness, hue-angle, chroma, etc.). To obtain the information about the perceptual consequences of the chosen transformations, perceptual image quality, naturalness, and colorfulness were measured by means of scaling. The results suggest that in general a more colorful image than the original one is preferred. Introduction Every color can be described in terms of hue, colorfulness (chromaticness), and brightness1. For an optimal reproduction of natural images on color displays it is important to know the influence theses perceptual attributes may have on perceived quality2 and naturalness of such images. We are interested in naturalness because it is expected that qualitatively optimal images appear, at least, natural. This prompts the investigation into the perception of images transformed in a “natural” way. Consequently, global attributes have to be considered. Changes in global colorfulness due to chroma variations were chosen as the first step in a series of investigations into possible transformations. Method: Image Transformation Digitized color images of four natural scenes were described by their color point distributions in the CIELUV uniform color space. For every original image a set of new images was computed by changing the chroma value for each pixel keeping lightness and hue-angle constant. Two methods of changing chroma were used: 1) translation: the chroma value was increased or decreased by adding/subtracting the same number of chroma units to/from the chroma value for every pixel; 2) multiplication: within each image the chroma value for every pixel was multiplied by a constant. If during the processing of the images calculated r, g, bgrey values were out of the possible range for the grey values of the monitor, the nearest possible value of chroma was used (clipping). For every scene, twenty images were prepared using ten different constants of translation and ten of multiplication. For every image, average chroma value and variability of chroma (standard deviation) were computed. 60 50 40 30 20 10 0 2 4 6 8 multiplication translation average chroma qu al ity F r u i t