FZ38 Resolution (part 2, contd)

 


Image Comparisons and Discussion


I compared on-screen various areas of the images side-by-side at the different ISO settings at 100% and 300% and made screen captures of these comparisons, which are shown and discussed below (click on the image to see the full version).



Crop comparison at 100%: Es chart and fuselage


For 100, 200 and 400 ISO in the above comparison, the coloured Es look fairly similar when viewed at 100%, but both the 800 and 1600 ISO crops show severe bleeding of the red Es and fading of other coloured Es, with cyan and yellow faded to very pale greys.  For the coloured fuselage areas in the lower part of each crop, while for 800 ISO there is some loss of detail and spreading of colours compared with the lower ISOs, it is only at 1600 ISO that loss of small details, smudging of colours and noise becomes very apparent, and even then the general shapes, lines and colours can be seen.



Crop comparison at 100%: grey scale, narrowing lines chart and fuselage


The above crop comparison is similar to the previous one, except that since the top part of each crop only contains grey/black, fading of colours is not evident.  The changes in the fuselage area in the lower part of the crop are similar to the previous comparison, except that at 800 and 1600 ISO, there is virtually complete loss of yellow colour from the thin tail-fin area.  The resolution as shown by the narrowing black lines seems to show a gradual trend of decreasing resolution with ISO, all the way from 100 ISO to 1600 ISO.



Crop comparison at 300%: coloured Es chart


At 300% this comparison shows that 100 ISO and 200 ISO look fairly similar, with the resolution at 200 ISO being only slightly less than at 100 ISO.  At 400 ISO, the resolution is only slightly less than at 200 ISO, but there is "bleaching" to grey of the coloured Es smaller than about 1.8, except for the red Es. At 400 ISO for the Es larger than 1.6 the bleaching gets progressively less severe as the Es become larger, but the trend for yellow is not regular, being patchy, and the yellow has a greenish tinge.  The black Es seem unaffected throughout.  For 800 and 1600 ISO all the coloured Es, even red, are "bleached" and smeared out to grey. While the red, blue and magenta Es are replaced by a grey of about the same intensity, the green, cyan and yellow Es have virtually faded away.


There is a gradual reduction in resolution from 100 ISO through 400 ISO when looking at the black and red Es, and this trend continues for the  black and "faded-to-grey" red Es for 800 and 1600 ISO.  The loss of resolution with ISO is much more severe for the other colours and worst for cyan and yellow, followed by green.




Crop comparison at 300%: front of plane fuselage


In this comparison there is a general trend of increasing noise and loss of detail with increasing ISO.  At 800 and 1600 ISO there is also obvious bleeding of red.  Looking at the wheel, at 100 ISO there are a couple of small faint yellow blotches, and with increasing ISO these increase in number and intensity.  At 800 and 1600 ISO there is also one small very bright yellow-green coloured blotch at the front of the wheel, and at 1600 ISO there is a similarly coloured blotch but in a different place, and it is larger and brighter.


Conclusions


Resolution and IQ

It seems that for the smaller coloured Es, when examined at 300%, the limitations of the Bayer filter/demosaicing process become apparent.  Although luminance information is collected at every pixel, only part of the colour information is collected at a given pixel, and the remainder needs to be interpolated from surrounding pixels, thus the color resolution is lower than the luminance resolution.  Obviously the worst colour resolution is for yellow.  When the ISO is too high (the signal-to-noise too low) it seems that there can be insufficient colour information to allow the demosaicing to represent the E as a colour, but still enough luminance information available to represent the E as a grey shade.


The outcome of this for the coloured Es is that the resolution varies with colour.  Considering only the black Es, the loss of resolution is only gradual as ISO increases.  The degradation in IQ with ISO in terms of increasing noise, loss of detail and degradation of colours seems to far outweigh the loss of chart resolution measured from the black Es or narrowing black lines.


To achieve the best possible resolution and IQ it can be concluded that it is necessary to keep the FZ35/38 at a setting of 80 or 100 ISO, although 200 ISO will give only a slight loss.


The Importance of Scale

It is important to consider the scale at which an image will be utilised and viewed when considering the above results.  Impressions of IQ alter with the scale of viewing and whether broad coloured areas or fine detail is viewed. 


As can be seen when viewing the full sized images of the entire chart board at 1000 x 750 pixels, the 400 ISO image looks quite good when compared with the 100 ISO image (which is very similar to the 200 ISO image), and even the 1600 ISO image does not look too bad.  If the lower ISO images were not available for direct comparison then the 1600 ISO image could be considered slightly noisy but otherwise OK.


Even when viewed at 100% (as shown in the 100% crops above), the 100, 200 and 400 ISO crops do not look very much different, and it is only at 800 ISO where the deterioration in viewed IQ becomes apparent.  It requires viewing at 300% (as in the 300% crops) to be able to see an obvious stepwise deterioration in IQ going from 100 ISO up to 1600 ISO.


Therefore, from a practical point of view, if images are not going to be cropped and are only ever going to be viewed on a computer screen at a reduced size, images taken at higher ISO settings may be perfectly adequate.  However, if planning to crop and/or print an image and the best IQ is needed then it would be beneficial to set the cam at 80 ISO or 100 ISO.




Return to FZ35/38 Resolution Testing Part 2


Return to Home Page