We have studied a new metric for color image noise (graininess) evaluation that is also applied to monochrome images. Graininess would be the most important and basic attributes for pictorial images, and Graininess Scale (
GS) has been widely used as a noise perception model. Similar to the treatment adopted in the
GS, we have computed Lightness Noise (
LN) by using the lightness instead of the reflection density and by using a new sensitivity function based on a model describing our observations. But
LN would not be sufficient for chromatic images as it is. So, we focused our attention on the fluctuations of both chroma and hue-angle on color images as these decrease perceived color image quality. Chromatic Noise (
CN) has computed by using these fluctuations with consideration to spatial filters by the human visual system. Finally, we have obtained Graininess Index (
GI) by using both
LN and
CN. To confirm our approach, we made subjective assessments for test patches, and got good relationship (γ=0.91) between
GI and subjective evaluated levels.
View full abstract