Software for Image Contrast Enhancement
- Created on Friday, 01 April 2011
This technique provides detail and edge contrast enhancement, latitude reduction, and noise reduction when applied to digitizers for X-ray imaging.
Whenever images are acquired with a large dynamic range, a problem arises in reproducing these images on a medium (i.e. CRT, paper). There are the restrictions of the recording apparatus (number of reproducible intensity levels, color gamut, and color resolution) as well as environmental factors (flare, light level).
Apart from these technical restrictions, there are also fundamental bounds to the capabilities of the human visual system. Our visual system is not able to discern the finest gray-level differences or color differences that can be acquired by the acquisition apparatus, and therefore details in the images are unperceived after rendering the image. The recording problem is therefore not only a question of how to map the acquired range onto the recording range, but how to arrange this mapping so that our visual system is able to resolve most of the details present in the acquired scene.
This dynamic range problem arises in digital radiography, where 12 bpp gray-level images are acquired, in scanning and digital photography, where 24 bpp or larger dynamic range color images are acquired. The more sensitive the acquisition devices become, the more need there is for an adequate solution to this problem.
The application of gradation curves can enhance the image in one part of the range, but only at the expense of losing details in other parts of the range. Similarly linear filtering (e.g. unsharp masking) can enhance the image in a certain spatial frequency band, but only at the expense of losing details at other spatial frequencies.
A multi-resolution approach, taking detail amplitude and not spatial extension of details as the fundamental entity, has a higher potential for image enhancement. Unlike radiographic images, these real world color images have an additional naturalness constraint. Image processing that gives the images an unnatural look cannot be tolerated.
Instead of the pixel values, this technology processes the image details at different resolution levels. At every resolution level, small amplitude details are enlarged while large amplitude details are minimized. In that manner, the available dynamic range is used in a smarter way.
This software makes all details visible in black and white, color, and medical digital images, independent of their size or contrast level. It decompresses the digital image in a great number of different frequency ranges, and modulates the signal amplitudes (or contrast) within each of these ranges. This results in detail contrast enhancement, edge contrast enhancement, latitude reduction, and noise reduction. Loss of frequencies that occurs in unsharp masking is avoided. The software can also be applied to color images.
Details are visualized better, independent of their size or contrast level. In contrast with other contrast enhancement methods such as unsharp masking, artifacts are avoided and a high-fidelity image is retained. In the case of color images, color naturalness is retained.
The software is commercialized in digitizers for medical X-ray imaging but can also be applied for other purposes, such as color image processing and improving image quality.
This technology is offered by AGFA-Gevaert NV. For more information, view the yet2.com TechPak at http://info.hotims.com/34453-194.