Fig 1. Analysis of an edge created using Photoshop, with segments blurred by Filter->Blur->Gaussian blur, with radius settings ranging from 1 to 28 px. The segment at the cursor was blurred with a radius of 12 px.
LoCA Focus Analyser uses a simple method to measure edge spread -- it looks along a column of pixels. MTF software, such as QuickMTF, use a method that aggregates information from many columns (for more info, google 'slanted edge spread function'). How do the two compare?
I created an image using Photoshop with edge segments having various degrees of Gaussian blur (Filter->Blur->Gaussian blur). Figure 2 shows the Focus Analyser report for the resulting edge. (Update: Similar image generation can be now done using this edge image synthesizer.)
Fig 1. Analysis of an edge created using Photoshop, with segments blurred by Filter->Blur->Gaussian blur, with radius settings ranging from 1 to 28 px. The segment at the cursor was blurred with a radius of 12 px.
Each section was also evaluated using QuickMTF. For example, figure 2 shows the QuickMTF edge spread analysis for the same section (marked by the cursor) in figure 1, where the Gaussian blur was radius 12:
Fig 2. QuickMTF analysis of an edge blurred by Photoshop's Gaussian blur with a radius setting of 12 px.
Figure 3 shows Focus Analyser and QuickMTF measurements at various blur levels. The first two blur levels (1 and 2 pixel radius) had sharpening artifacts, probably introduced by Photoshop during rotation or resizing.
Fig 3. Comparison of edge spread measurements by Focus Analyser and QuickMTF for the synthesized blurred edge segments in figure 1. QuickMTF's spreads are about 12% larger than Focus Analyser's.
QuickMTF's edge spread measurements are generally ~12% larger than Focus Analyser's for these synthesized edges.
Figure 4 is the Focus Analyser report for a slightly defocused, perpendicular (not inclined) razor blade edge. A defocused image was chosen to provide a wider edge spread.
Fig 4. Focus Analyser report for a razor edge, photographed perpendicularly. Chromatic aberration caused the red channel to be less focused than the blue and green channels.
Figure 5 shows the QuickMTF analysis of the central region:
Fig 5. QuickMTF analysis of the central area of the edge in figure 4. The spread for the red, green, and blue channels was found by QuickMTF to be 10.4, 7.6, and 7.4 px respectively, about 10% larger than the spread reported by Focus Analyser.
In the central region, Focus Analyser says the blue and green channels have an edge spread of approximately 6.7 px (as shown in figure 4) and the red channel has a spread of 9.6 px. QuickMTF says blue and green have a spread of 7.6 and 7.4 px respectively, and red has a spread of 10.4 px. That's a ratio of 112% and 108%, which is comparable to the ratios found using synthetized edge spreads.
An MTF50 value is based upon an analysis of an edge spread, so the two are likely related for a given edge shape. Figure 6 shows how MTF50 of Gaussian blurs measured above varied with edge spread, as reported by QuickMTF and Focus Analyser. The two right-most columns repeat the QuickMTF edge spread and MTF50 figures in units of LP/PH (line pair per picture height), often used.
Fig 4. Comparison of MTF50 and edge spread measurements by QuickMTF and Focus Analyser for the synthesized blurred edge segments in figure 1.