High Pass Filters


Today’s Question: Canon has introduced a new 5DS R, a camera without a high pass filter over the sensor. This is supposed to make the image sharper. In Photoshop we use the high pass filter to increase sharpness. Please explain.

Tim’s Quick Answer: The short answer here is that the High Pass filter (in Photoshop for example) doesn’t truly sharpen a photo. Rather, it removes low detail areas of an image and retains high detail areas, creating contrast in the process. That contrast can then be used to sharpen with the help of a blend mode. A high pass filter on a camera’s image sensor filters out some of the low frequency information to avoid artifacts, but softens the image in the process of filtering out some of the information projected by the lens.

More Detail: In other words, both filters are doing approximately the same thing, but they are generally used in different ways. A high pass filter will filter out low frequency information and retain high frequency information. This is helpful in image sensors for preventing moire patterns and other visible artifacts in a photo. However, because some information is being filtered out, there is a degree of softness imparted to the image. In other words, there is a tradeoff here, but many photographers would choose a higher degree of sharpness and detail over the potential for removing certain artifacts in the image.

The High Pass filter in Photoshop also filters out low frequency (low detail) information in a photo, preserving (and enhancing) the high frequency (high detail) areas. The result is something of an embossed effect. When combined with one of the “contrast” blend modes (such as Overlay), this can create a sharpening effect. The light areas of the embossed copy of the photo will brighten one side of a contrast edge, and the dark areas of the embossed copy will darken the other side, resulting in the appearance of greater detail and sharpness.

So, in both cases the same basic filtering of information is being used for different purposes, but with variations on how those filters are implemented contributing to the difference.