Image Production · Digital imaging and processing2 min read5-question check
Digital Image Processing: Corrections, Enhancement and Stitching
Corrections fix the detector, edge enhancement sharpens, smoothing quiets noise, and equalization shows thick and thin parts at once.
Detector corrections
Before an image is shaped for viewing, the raw detector signal is cleaned up. The analog-to-digital converter turns each detector element's signal into a number, a step called quantization, and the bit depth sets how many shades of gray are possible. Flat-field correction, also called gain and offset calibration, evens out small differences in how each element responds, so a uniform exposure produces a uniform image. Dead pixel correction fills in elements that no longer respond by averaging their neighbors. These corrections depend on detector calibration, which is why calibration is part of routine QC. After that come histogram analysis, rescaling to a standard brightness, and the look-up table that sets contrast for the body part.
Spatial frequency processing
Spatial frequency processing changes how edges and fine detail look. Small structures and sharp edges are high frequencies. Edge enhancement boosts them, so fractures, trabeculae, and lines look crisper, but it also boosts noise. Too much creates a dark or bright halo next to high-contrast edges, such as a hip prosthesis, and that halo can mimic loosening. Smoothing does the opposite. It averages neighboring pixels to quiet noise, at the cost of blurring fine detail.
Equalization
Equalization, sometimes called dynamic range compression, lifts detail in very bright and very dark areas at the same time. A chest shows lung markings and the mediastinum together. A lateral C-spine shows the dense shoulders and the thin upper spine together. It compresses the range of densities instead of adding exposure.
Stitching and limits
Stitching joins several overlapping exposures into one long image for scoliosis series and long leg or full spine measurements. A radiopaque ruler or calibration marker helps the software line up the images and allows accurate measurement. The patient must not move between exposures, or the joined image will show steps, gaps, or doubled anatomy. Processing makes images look consistent, but it cannot add information. Rescaling can make an underexposed image look the right brightness while the quantum noise remains. It cannot restore data lost to detector saturation either. Getting the exposure right comes first.
Clinical Example
A radiologist asks why a hip arthroplasty follow-up shows a thin dark band around the femoral stem that was not on the last study. The radiographer sees the exam was processed with a high edge-enhancement setting. Reprocessed with the standard hip algorithm, the band disappears, and the stem is well seated.
Common Mistakes
- Using edge enhancement to fix a noisy image; it makes noise worse.
- Expecting processing to rescue an underexposed or saturated image.
- Letting a scoliosis patient shift between stitched exposures.
- Confusing flat-field correction with window level adjustments.
Why It Matters
Image processing is listed by name under Equipment Operation for 2027, including flat fielding, dead pixel correction, edge enhancement, equalization, smoothing, and stitching. Expect questions that describe an image appearance and ask which processing step caused or fixes it.
Key Relationships
- Flat-field correction → uniform detector response
- Dead pixel correction → interpolate from neighbors
- Edge enhancement → sharper edges, more noise, possible halo
- Smoothing → less noise, less detail
- Equalization → thick and thin areas visible together