Digital cameras — Resolution and spatial frequency responses

Standard Background and Technological Evolution

ISO 12233:2024 is the international standard for measuring the resolution and spatial frequency response of digital cameras, revised by ISO/TC 42 Photographic Technology Committee, replacing ISO 12233:2023 (fourth edition). Since its first publication in 2000, this standard has undergone five iterations, gradually evolving from a single visual resolution test to a multi-dimensional spatial frequency response (SFR) measurement system. With the increasing complexity of image processing algorithms in digital cameras (such as adaptive tone mapping and compression), the standard introduced low-contrast edges and a sine wave star chart test chart to reduce the impact of nonlinear processing on the measurement results.

Core Terminology and Definitions

The standard defines 31 key terms, among which the core ones include:

• Spatial Frequency Response (SFR): The relative response of the output amplitude of an imaging system to changes in the input spatial frequency, typically normalized to 1 at zero frequency.
• Visual Resolution: The highest spatial frequency at which an observer can distinguish black and white lines, expressed as linewidth per image height (LW/PH).
• Edge-Based Spatial Frequency Response (e-SFR): The SFR calculated using tilted edge analysis, applicable to horizontal, vertical, and diagonal directions.
• Sine-Based Spatial Frequency Response (s-SFR): The SFR measured using a Siemens star map (sine modulation), which is more robust to nonlinear processing.

Test Conditions and Requirements

Visual Resolution Measurement Method

The standard recommends using the CIPA resolution test chart (Figures A.2-A.5), containing a hyperbolic wedge pattern, with a spatial frequency range of 200-2500 LW/PH. Subjective judgment rule: start from low frequencies, and only when all lower frequencies are clearly visible is it considered "resolved". Computer-aided software (such as HYRes) can automatically analyze, and the results are consistent with subjective judgment. This method is suitable for rapid inspection on end-to-end production lines.

Edge Base Spatial Frequency Response (e-SFR)

e-SFR uses a four-leaf tilted star chart test chart (Figure 4), with an edge reflectivity ratio of 4:1 (preferred) or 6:1 (optional).

1. Select edge region (ROI, 100-400 pixels wide)
2. Inverse OECF linearization of image data
3. Calculate one-dimensional derivative and window (Tukey window, α=0.5)
4. Calculate edge centroid position row by row
5. Fit edge curve with fifth-order polynomial (linear optional)
6. Form oversampled edge spread function through shifting and binning (4x oversampling)
7. Differentially obtain line spread function, calculate DFT amplitude after windowing
8. Correct the differential filter response (inverse sinc function)

Key improvement: Use polynomial fitting to compensate for geometric distortion (such as wide-angle lens), avoiding the error of traditional linear fitting. Non-uniform illumination can be compensated by the algorithm in Appendix J.

Sine-Based Spatial Frequency Response (s-SFR)

s-SFR uses a Siemens star map (Figure 6), sinusoidal modulation, and a reflectivity ratio of 50-250:1. Analysis algorithm (Appendix F): The star map is divided into 24 angular segments, pixel values are extracted radially and fitted with a sine curve, and the modulation index is calculated (M = (Imax - Imin)/(Imax + Imin)). Normalization is performed by dividing by the modulation index of the brightest and darkest gray blocks. s-SFR is insensitive to edge detection-type image processing and is suitable for evaluating the effects of nonlinear processing. At least four images are typically taken and averaged.

Result Presentation and Reporting

Standard output format:

• Resolution: expressed in LW/PH, can report basic (isotropically isolated) or representative (minimum or average) values.

• SFR Curve: Presented graphically (Figure 7) or in a list format, with frequency units of period per pixel, LW/PH, or period per millimeter. Half-sampling frequency regions must be labeled.
• s-SFR Polar Plot: Shows the modulation level at each angle (Figure 8), recommended for multi-directional analysis.

Also, report all camera settings that affect the measurement results (lens focal length, aperture, compression mode, sharpening settings, etc.).

Implementation Recommendations

In practical applications, it is recommended that:

• Preferably use a 4:1 reflectance ratio test chart to avoid data cropping leading to an overestimation of SFR.
• For cameras with strong nonlinear processing (such as smartphones), s-SFR is more stable than e-SFR.
• Combine the two methods to evaluate the impact of image processing on SFR.
Strictly adhere to illumination uniformity and focusing conditions; otherwise, use non-uniform illumination compensation. Include all variables in the report to ensure reproducibility of results. This standard provides a complete scientific framework for evaluating digital camera resolution, effectively supporting product quality control and consumer selection. The measurement method may further evolve in the future as AI imaging algorithms develop.