General image enhancement framework

DETAILED DESCRIPTION FIG.
1 illustrates an embodiment of the invention.
According to FIG.
1, approximator 11 receives an input image and generates at its output an approximation of the image.
The approximator 11 may be in anticipation of the image enhancement to be performed.
For example, in the case of noise reduction image enhancement, approximator 11 may incorporate a low pass filter that filters the input image for high frequency image content of which noise is often a primary part.
In another example, approximator 11 may be one that applies a moving window average on the input image to generate a "smoother" (less jogged) version of the input image.
Approximator should be designed to generate a version of the image that has a higher likelihood of containing visually important image details that should not be distorted or misinterpreted during image enhancement.
With the approximated image thus generated the input image and the approximated image are supplied to differential mapping 13 that uses the images to form a pre-enhancement map.
For instance, in one embodiment, differential mapping 13 may incorporate a subtractor that subtracts the approximated image (pixel-by-pixel) from the input image to form the pre-enhancement map.
The pre-enhancement map comprises differential information of the input image and the approximated image thus eliminating features common to both.
In the case of noise reduction image enhancement, the pre-enhancement map may include some true image information along with noise that is to be reduced.
the level of true image information other than noise contained in that pre-enhancement map would thus be dependent on the effectiveness of approximator.
For an ideal approximator, the differential mapping mainly yields the noise present in the image.
Since this design would be typically known, the design factors may be taken into consideration when parameterizing (such as setting threshold values for edge, noise discrimination).
An enhancement engine 15 receives the pre-enhancement map at its input to process the map into an enhanced map