HTI5203 Chap.8 Resolution, Noise and Modulation Transfer
Resolution, Noise and Modulation Transfer
The official Week 8 block groups resolution, modulation transfer function, scatter, noise, artefact and dynamic range. This chapter keeps those limitations distinct. It reads MTF at the spatial frequency relevant to an imaging task, separates fine-detail transfer from random variation, and diagnoses fixed, random and directional patterns before selecting a correction.
It avoids inventing detector thresholds absent from the available course files. A system comparison is task-dependent. MTF must be read at a named spatial frequency, and crossed curves can reverse the ranking at another detail scale. Random fluctuation, broad contrast loss, fixed detector patterns and directional blur point toward different mechanisms and require different checks.
More exposure may address quantum noise but cannot restore information lost through motion, geometry or processing. The defensible sequence is observation, named property, proposed mechanism, discriminating test and bounded correction.
What this chapter covers
- 01
Spatial resolution
- 02
Modulation transfer function
- 03
Spatial frequency
- 04
Quantum noise
- 05
Scatter and contrast
- 06
Fixed and directional artefacts
- 07
Dynamic range
- 08
Task-specific comparison and mechanism-matched correction
Worked example · free
Compare crossed MTF curves
- 2Compare at low frequency.
- 2Compare at the finer frequency.
- 4Relate the ranking to the imaging task.
Key terms
- Spatial Resolution
- The ability to distinguish small or closely spaced image details.
- Modulation Transfer Function
- A function describing retained image modulation across spatial frequency.
- Noise
- Random variation that obscures useful signal.
- Artefact
- A structured image feature that does not belong to the imaged object.
Resolution, Noise and Modulation Transfer FAQ
Is a smooth image always high quality?
No. Processing can suppress visible noise while also smoothing fine detail. Image quality is multidimensional, so resolution, noise, contrast, scatter, artefact and geometry must be evaluated against the task rather than collapsed into appearance.
How should I compare two MTF curves?
Name the axes, identify the spatial frequency relevant to the task and read each curve there. A higher transfer at that frequency indicates better preservation of that pattern. Curves can cross, so one global ranking may be misleading.
How can I distinguish an artefact from noise?
Noise is random and changes between images. A structured line, patch or band that repeats in detector coordinates suggests artefact. Motion tends to produce directional blur related to the anatomy or acquisition rather than a fixed receptor position.
Exam move
Create one card each for resolution, noise, scatter and artefact. Record appearance, mechanism and a discriminating check. Practise reading MTF at a named frequency and confirm current units, criteria and receptor conventions in the Week 8 Canvas@PolyU material. Begin by refusing the label “poor quality” until a property has been named.
Make separate entries for spatial resolution, contrast, noise, scatter, artefact and geometric distortion. Give each entry an observable pattern, a plausible mechanism and one observation that could disprove it.
Fine structures merging while the background remains smooth points toward resolution, random fluctuation points toward noise, broad contrast loss may suggest scatter, and a feature fixed in detector coordinates suggests an artefact. These are diagnostic starting points, not automatic conclusions, so every card needs a discriminating check. For MTF, draw axes before drawing a curve.
Put spatial frequency on the horizontal axis and modulation transfer on the vertical axis, then explain why retained modulation commonly declines as detail becomes finer. Choose two frequencies and compare two invented systems at both. If the curves cross, state that the ranking depends on the imaging task; do not crown one system universally better.
Practise reading a value at a named frequency and translating it into a statement about contrast transfer for structures of that scale. Confirm the exact units and any criterion in current teaching materials before using them. Build correction pairs that match the mechanism. Increased exposure may reduce quantum noise under appropriate conditions, but it cannot restore detail lost to motion, geometry or detector response.
Collimation or other scatter control may improve contrast but has technique implications. Processing can make noise less visible while smoothing useful high-frequency detail. For each intervention, write the intended property, the possible trade-off and a measurement that could verify improvement. This prevents a generic recommendation from being attached to every unsatisfactory image. End with a signal-versus-pattern drill.
Describe whether a defect is random or repeatable, local or global, directional or non-directional, and fixed to anatomy or detector coordinates. Only then select a mechanism. Compose short answers in four parts: observed feature, named property, proposed cause and discriminating check. When the evidence is incomplete, say what additional observation is needed.
Log wrong answers as property confusion, axis-reading error, task-free MTF ranking or mechanism-treatment mismatch. Use the Week 8 Canvas@PolyU material to confirm detector-specific conventions and thresholds rather than borrowing them from another system.
