Thursday, 23 October 2014

Infrared photography with a point-and-shoot Panasonic Lumix DMC-FH27

On trying digital infrared photography for the first time I bought on eBay a simple IR-converted point-and-shoot camera, a Panasonic Lumix DMC-FH27. In some ways it was an easy and fairly cheap means of seeing what was possible. For example, custom white balance is very easy to set. I found a lens hotspot at the wide end (5 mm) of the zoom but that disappeared at longer focal lengths). The biggest problem is noise. Within the sky or clouds, there is very marked noise even at the lowest ISO settings. In fact, the noise is so bad that attempts to change the hue of a particular colour chosen by the eye-dropper tool often failed and all efforts to apply colour control points in Nik software failed miserably. Moreover, the colour rendition of foliage was not what I had expected.

I should, of course, have read the Kolari Vision website before I bought it. This is what it says:

Most Panasonic cameras take photos with a more saturated, distinctive blue color. Their flash tends to mess up the custom white balance, and some high resolution sensors can have a distinct noise pattern. Another thing to note with these cameras is that the 590nm and 665nm filters tend to come out stranger, with the leaves being more red than yellow, and skies becoming somewhat green.

I then moved on to IR-converted DSLRs and my FH27 has been put aside with the hope it would fit in my pocket when shooting video or normal photographs so that I did not have to carry several heavy camera bodies.


A couple of weeks ago I it caught my eye and I wondered if I could find some way of overcoming the processing problems I had encountered earlier. I thought I would try, as a first step, to reduce the noise. This I did using Nik’s Dfine 2 (Aperture plug-in), first applying the basic, global adjustment and then using control points to either increase or decrease the noise (the latter when I wanted to retain detail). Then, I returned to Aperture and took the image to Photoshop Elements/Elements+ to swap the red and blue channels. I also added a hue/saturation adjustment layer to shift the overall hue as necessary. Then back to Aperture and Viveza 2 where I hoped I would be able to use control points as well as global levels and curves adjustment. I could and was, therefore, able to process images to achieve the ‘look’ I wanted, using the full panoply of controls available. A return to Aperture and then a visit to Sharpener Pro 3 where I applied only sharpening to control points where I wanted to see crisper detail. Global sharpening was not applied in order to retain the noise reduction I had obtained at the first step.
Screen grab from the loupe in Dfine 2 showing before
and after global noise reduction
That is the workflow or standard protocol I am applying to the jpgs from the FH27 (no raws on this camera alas).

Some of the older websites describing infrared techniques I found did suggest that for some small cameras noise reduction might be necessary to satisfy those photographers who did not think noise added to the atmosphere of infrared images. They suggested Neat Image, for example, which I have used to good effect for scanned prints on textured paper (post 22 December 2013). However, noise reduction software that offers local control rather than a simple global coverage is of a great advantage for the present application. Dfine 2, offers local control points and colour ranges to obtain selective noise reduction. By using Dfine 2 for the first step, I have been able, after that, to follow my standard workflow used for images from my Nikon DSLRs (16 October 2014).

A before and after sky area from a processed image showing the noise reduction
So would I recommend the FH27 as an infrared-converted camera? I cannot say that I would. A converted Nikon D70 or D80 body can be bought for not that much more and the small sensor with a large number of pixels (the pixel density of the Lumix is 24x that of the Nikon D80) is probably just too small to get really good low-noise IR photographs. Exposing at the red-end of the spectrum exacerbates noise for the reasons explained here. However, I am now able to say that shots can be processed to get much better results than might be expected from my early efforts and the amount of noise generated.

A colour version of a test image manipulated using control points in Viveza 2
A black-and-white conversion of the same image in Silver Efex Pro
The jpg produced by the camera
A quick shot between showers last weekend with white balance set on grass in sunlight a few seconds before
I could not make local adjustments with this level of noise
After global and local adjustments in Define 2, Viveza 2 and Sharpener Pro 3
The enlarged images of noise before and after shown above are from these images
Ignore the subject (I didn't want to get wet getting something more photogenic and the manipulation achieved. This is just to show the sort of effect I can now get from an almost infinite range of global and local adjustments