The Holding Tank

Home Page Photography

Photography RSS 2.0 Feed

| Photos & Themes RSS | Hardware | Technical | Links | Books and Magazines  | Downloads |

Technical

| Flash Tests | RAW Converter Tests | Depth of Field, Image Quality and Sensor Size |

| MTF | Resolution | Cropping and Printing | Image Stabilization and Magnification |

| Lens Image Simulation | Anti-Alias Filter and Sharpening | | Tilt and Shift Lenses |

Lens Distortion Correction and Image Detail | Digital Darkroom Workflow | Optical Coatings |

Effect of the Anti-Alias Filter and Sharpening

This is surprisingly difficult to obtain information on, after considerable searching I discovered this document produced by KODAK. This provides an MTF plot for quartz, plastic and no anti-alias filter normalised to the sample rate. What is not clear is what the no anti alias filter trace includes, the sensor, sensor and lens?

On the basis that the laws of physics and intelligence of designers are the same for Kodak and Canon, I am making the assumption that these results apply to Canon at least approximately.

In the figure below I have reproduced the plot rescaled for the 20D sample rate, so now the x-axis is in cycles per mm (I treat this as equivalent with line pairs per mm for simplicity, although this may not be exactly correct).

My assumption is that the Canon anti-alias or blur filter and sensor MTF follows the quartz plot. Comparing this with some plots of film MTF in the same log log axis format suggests the overall sensor and filter MTF is comparable to good quality film (plots available from Norman Koran’s web).

What is interesting, is does the effect of the filter and sensor dominate image quality, or is the additional degradation of a sharp lens detectable in a photographic type scene. Here the same method for the Lens Simulation of cropped image is used, the lens case being the 135mm f2L at 13mm offset from the optical axis. Please read the suggestions for examining sharpness. The Matlab code for these simulations is also in the Lens Simulation code archive.

Test Image

Test Pattern

The sharpening recommended, at least as a starting point, by Canon with the Photoshop Unsharp Mask is Amount: 300%, Radius: 0.3 pixels, Threshold: 0 pixels. (Ref “GETTING THE MOST FROM YOUR EOS-1 CLASS DIGITAL SLR”).

Spectral Performance

 The following plots show the amplitude of various line frequencies in the above photographic test images power summed over all directions and all three colour planes.

The anti-alias filter can be seen to be causing roll-off of the finer (higher frequency) image detail from about 10 cycles per mm (red trace compared to blue trace). The Canon recommended Sharpening rather overcorrects this loss up to about 50 cycles per mm (green trace).  If some additional blurring from a lens is added however, the recommended sharpening largely restores the detail level accurately up to about 35 cycles per mm.

The effects can be more easily seen by plotting these data relative to the original image line spectrum, below.

This shows that for the black trace which is fairly neutrally sharpened in the region of important detail of 10-40 cycles per mm, falls to about 50% at 60 cycles per mm. This is about 77% of the Nyquist frequency. This indicates the loss of the capture system when a balanced level of sharpening is performed and corresponds to the capture plus sharpening 50% mtf level of about 60 cycles per mm.

Turning attention now to the spectra of the test pattern of 10 and 30 lp/mm and looking in detail at that frequency range we can see the recommended sharpening of the anti-alias filtered pattern restores the contrast of the test pattern very well.

However close examination of the sharpened pattern shows that the spreading nature of the anti-alias filter followed by the application of the sharpening algorithm leaves sharpening artifacts for about 13 pixels around the line edges. This are only visible due to the very high contrast of the image. See detail below. These artifacts have a magnitude of about 10 (on a scale of 0-255) in the 10 lp/mm area, and up to 25 in the 30 lp/mm area.

Conclusions

Examining the resulting simulated crops shows there is detectible degradation between the Quartz Filter and the Quartz Filter plus EF 135mm f/2L @ f8 and 13mm off. Thus top lens quality is still worth striving for.

The effect of the recommend sharpening does a very effective job of clearing all but the most fine detail losses of the anti-alias filter with the minimum artifacts. If the lens is good and sharp this is probably all the sharpening that is needed. For softer lenses a little more sharpening will be beneficial.  

I have evaluated the sharpening of various RAW converters in RAW Converter Tests

Last Updated 05/06/2008

All Content © 2005-12 Lester Wareham All Rights Reserved     

All material is supplied  as is and without warranty, use at your own risk.  

All opinions stated are the authors own.

All quoted information remains the copyright of the respective authors.

Contact: 

Note this email gets spammed a lot so if you want to get your message through please prefix "Holding Tank" in the title

holdingtank@ware.myzen.co.uk