Dpreview's OM-D E-M5 Review, Second thread

Started Apr 30, 2012 | Discussions thread
J C Brown
Senior MemberPosts: 1,372
Re: Sensor Resolution - for R Butler et al.
In reply to Detail Man, May 11, 2012

Detail Man wrote:

J C Brown wrote:

My attention was drawn to this thread by Detail Man who kindly referred to my FZ50 paper in his response to Kikl. From the range of comments on this topic it is clear that there is some concern about the subjective interpretation of the test images obtained from the DPR test chart and its effect on the accuracy of the resolution measurements.

An excellent and a very informative post, Jimmy ! The in-camera sharpening that in-camera JPG "engines" perform tends to (artifically, using edge-enhancement techniques) "disguise" such things

Although intended for use only by private individuals I believe that the use of my test chart by DP Review to compare the colour resolution of cameras would be of interest to DPR members.

This is a truly useful test-chart. I wonder if DPReview, or any E-M5 owners/users, will endeavor to utilize your test-chart in order to assess it's "chroma-resolution" (as opposed to "luma-resolution") ?

Thanks for your very complimentary remarks Detail Man.

For the benefit of anyone who, suitably encouraged by Detail Man’s recommendations, might be interested in experimenting with my test chart, I feel that I should provide some more detailed information about its design, use and potential benefits.

The coloured Es chart is similar in design to the visual acuity charts used by opticians in that it uses rows of letters of different sizes to assess the limit of resolution but differs in that the letters are all Es, the incremental increase in size from row to row is just 10% of the size of the bottom row and to allow the resolution to be assessed for a range of colours the Es are printed in red, green, blue, black, magenta, cyan and yellow.

To measure the vertical resolution in lines per picture height, LPH, the distance between the camera and the chart is carefully adjusted until the length of the scale line in the recorded image is 400 pixels within one or two per cent. Then with the white balance set manually a shot is taken for each desired condition, e.g. aperture, ISO rating, etc.

As the thickness of the lines in the bottom row of Es corresponds to the height of a single pixel the vertical resolution for each colour can be calculated by dividing the number of pixels in the height of the sensor by the thickness in pixels of the lines of the smallest E of each colour which can be easily recognised as consisting of three lines and two spaces.

For example for a 12 MP sensor with a sensor height of 3000 pixels, if the smallest easily recognisable blue E has a line thickness of 1.5 pixels then the vertical resolution for blue is 2000 LPH.

As illustrated in the following image when examined at high magnification, by clicking on the image, the individual pixels which make up the lines of the Es can be very easily recognised allowing the resolution limit to be determined with acceptable accuracy.

In addition to the benefit of allowing the resolution to be assessed in six colours as well as black, experience with several users has shown that the consistency with which the resolution can be assessed using the coloured Es chart is considerably better than for charts consisting of tapered black lines for which the ability to locate the position at which it is possible to distinguish 9 lines and 8 spaces is much more varied and subjective. The edge and corner resolution can also be easily assessed by placing additional copies of the test chart in the appropriate positions on the test board.

As can be seen from the above Panasonic FZ50 image there is a significant variation in the resolution for different colours, with the green, cyan and yellow Es being considerably less well defined. These differences may be explained in part by the arrangement of the red, green and blue filters in the Bayer matrix and the in camera demozaicing process and in part by the effect of the edges of the image partially overlapping adjacent pixels.

To demonstrate that effect I designed the stepped Es test chart shown above. In that chart each successive E in a row is displaced upwards by 10% of the thickness of the lines of which the Es are constructed. When an image of the chart is recorded from the distance at which the thickness of the lines in the smallest E is one pixel, each successive E in each row of Es in the resulting image will overlap the pixels on the sensor by an increment of 10% of the thickness of the lines used to construct them.

The above images show part of the test chart and the corresponding part of an FZ50 image. As may be readily seen by clicking on these images the accuracy with which successive Es are recorded and the ease with which they can be recognised varies with the amount by which the edges of the lines overlap the adjacent pixels.

These results lead me to conclude that the appearance of images recorded using charts constructed of tapered black lines is susceptible to variation due to the overlapping pixel effect described above making it more difficult to obtain consistent measurements of the limit of resolution.

In addition to providing images for which the resolution is relatively easy to assess, the colour test chart gives the advantage of allowing the resolution to be measured for six colours and of facilitating assessment of the variation of colour resolution with various settings such as aperture and ISO rating.


-- hide signature --

J C Brown

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow