Overview
Image Quality Assessment (IQA) refers to the subjective and objective analysis and evaluation of the imaging performance (such as resolution, noise, distortion, color, dynamic range, etc.) of an imaging system under given conditions (such as the captured scene, lighting environment, and display device input of a camera). Image quality directly affects the practical value of an image—for example, the visual experience of human observers (such as appreciating photos or videos) or visual effects (such as providing indirect vision using Camera Monitor Systems (CMS) for road vehicles), as well as the target recognition rate of machine vision (such as the diagnostic accuracy of medical imaging, the safety of ADAS (Advanced Driver Assistance Systems), etc.).

Taking camera image quality assessment as an example, the assessment dimensions can be described through image attributes:

• Global attributes: Insensitive to changes in viewing conditions, unaffected by magnification, distance, or size, such as geometric distortion (lens barrel / pincushion distortion), dynamic range (ability to retain details in highlights and shadows), luminance uniformity (consistency of illuminance between the image edges and center), etc.
  
• Local attributes: Directly affected by the parameters of the imaging device and vary significantly with viewing conditions, such as sharpness (MTF values), noise level (Signal-to-Noise Ratio, SNR), color reproduction capability, etc.
  

Subjective and Objective Image Quality Assessment Methods
Subjective Assessment
Refers to obtaining quality scores through standardized procedures based on the visual perception of human observers. The assessment focuses on the intuitive visual impression of the image, including the overall clarity and transparency of the picture, the comfort of the color style (such as whether there is a color cast and whether the saturation is natural), the dynamic range performance (visual balance of details in highlights and shadows), the impact of noise visibility and distribution on the visual experience, and the image texture brought by sharpness.

Common Subjective Assessment Methods:
Absolute Assessment: Observers refer to the original image and score on a 5-point scale (1 = poor, 5 = excellent), outputting the Mean Opinion Score (MOS).
Relative Assessment: Without reference to the original image, observers compare the quality order of a batch of images, outputting the Differential Mean Opinion Score (DMOS).

Objective Assessment
Quantifies image quality through mathematical models, divided into three categories:
Full-Reference (FR): Comprehensively compares the differences between the distorted image and the original reference image to quantify quality loss. For example, calculating color difference based on a 24-color standard color chart to evaluate the color reproduction capability of the imaging system.
Reduced-Reference (RR): Extracts only key features of the original reference image and compares them with the corresponding features of the distorted image.
No-Reference (NR): Directly analyzes the features of the distorted image without the original reference image. For example, capturing a uniform light source target to test the luminance uniformity of the imaging system.

Among the three categories, Full-Reference objective assessment has become the mainstream due to its high accuracy and adapts to testing needs in multiple fields (such as camera测试用例 / adas测试用例 / cms测试用例, etc.). Taking the objective assessment of digital camera image quality as an example, its core application logic is: in a given lighting environment (such as a darkroom, D65 light source), simulate real scenes through diverse test targets (such as ISO 12233 resolution chart, standard color chart) and light source conditions (such as low illuminance), use the imaging system to capture the test targets to obtain distorted images, and then calculate the images through image quality analysis software (such as RIQA Image Quality Analysis Software) to obtain the image quality analysis results.

Common Image Quality Assessment Tools (Taking Digital Cameras as an Example)
Objective Assessment Tools

• Test Charts: Such as SFR slanted-edge test chart (for measuring Spatial Frequency Response), 24-color standard chart (for measuring white balance), grayscale test chart (for measuring dynamic range), checkerboard (for measuring distortion), etc. Chart types are divided into transmissive and reflective, mostly using light-resistant materials to ensure pattern accuracy;
  
• Light Sources: Including transmissive (such as transmissive light booth) and reflective (such as multi-color temperature standard light source booth), which can simulate test environments with different color temperatures and illuminance conditions;
  
• Optical and Acquisition Equipment: Collimator (simulates infinity scenes), image acquisition card (high-precision acquisition of image data, ensuring no loss of original signals);
  
• Analysis Software: Such as RIQA Camera, etc., which can calculate quantitative metrics such as MTF (Modulation Transfer Function), SNR (Signal-to-Noise Ratio), and ΔE (color difference) based on test samples, supporting model calculations for full-reference, reduced-reference, and no-reference objective assessment methods.
  

Subjective Assessment Tools

• Lighting Equipment: Covering transmissive (such as transmissive light booths) and reflective (such as multi-color temperature standard light source booths), which can simulate real environments under different color temperature and illuminance conditions;
  
• Real-scene Props: Such as plush toys (for texture and detail level assessment), humanoid mannequins (for skin tone reproduction testing), multi-color crayons (for color saturation and color cast assessment), standard graphic traffic lights (for overexposure and color recognition verification in high-brightness areas), simulated license plates (for character recognition and reflection suppression capability testing), etc.