Digital IR - Canon 5D MKII
I have been using infrared light to make images since 1995, and have worked with digital SLR cameras for infrared photography since the end of 2004. I currently use a Canon 5D MKII, converted by Lifepixel in the USA for my infrared photography (since April 2009). This is the fourth digital camera I have used for infrared photography (previously I used a Nikon D80 (2006-8), a Nikon D70 (2005-6) and a Sigma SD10 (2004-5). From 1995-2003, all my infrared photography was done on infrared film (35mm and 120).
Canon 5D MK II
In April, 2009 my second 5D MKII body arrived back from conversion, and I began using it for my infrared photography. While the primary motivation for the upgrade was image resolution (and 21mp is a LOT more resolution than 10mp), there were other unexpected advantages to the upgrade. Infrared light focuses to a different distance than visible light, rendering most AF system inaccurate with IR bodies. With the Nikon D70 and D80, it was possible to correct the AF system for a single lens, but auto focus with other lenses were rendered less accurate by this adjustments. A big surprise with the Canon 5D MKII was that no lens had accurate autofocus. I did some experimentation with the focus compensation, but had no sucess. Fortunatly, the 5D MKII permits manual focus to ber used when in LiveView, and even better, it permits an image to be magnified to 100% resolution for more accuracy in manual focus. This dramatically changes the accuracy of focusing with the IR camera, and permits working with any lens aperture, without any fear of focus issues. A further advantage of working in LiveView is that it provides a preview of the image exposure, with a live histogram, which makes exposure wirth infrared easy to do (as opposed to based upon experiance and testing, which was the approach used with my pervious cameras.
The Spectrum of Light
The full spectrum of light is broken up into three distinct bands, only a portion on which is normally visible. Ultraviolet light (UV) has the shortest wavelength, and is invisible to the human eye. Visible light is what the human eye sees normally, and ranges roughly from 400nm (violet) to 700nm (red). Infrared light (IR) has the longest wavelength and is also invisible to the human eye
This is a spectral sensitivity graph for a generic photographic CCD sensor, similar to that used on the Canon EOS 5D MKII (the specific spectral sensitivity of the sensor in the EOS 5D MKII has not been published to my knowledge).
The visible spectrum only uses about 60% of the sensor’s sensitivity, with ultraviolet and infrared light making up the other 40% approximately. Compared to Kodak’s HIE film, the incredible sensitivity of CCDs to infrared light becomes apparent.
An infrared longpass filter is designed to block all but the longest wavelengths, essentially blocking visible light while passing infrared light. This graphic shows the light transmitted by the filter installed in the Canon EOS 5D MKII by Lifepixel. It has a much steeper cut-off rate than the filter I’d had installed in the Nikon D70 in 2005, which leads to a slightly different look to the resulting images.
Normally, DSLR cameras have a highpass filter installed over the sensor, which blocks all the long wavelengths of light (which happen to be the infrared wavelengths).
The combination of the infrared sensitivity of the sensor and the steep cut-off curve of the Lifepixel infrared longpass filter’s transmission leads to the perfect situation for infrared photography.
The extended infrared sensitivity of the sensor is taken advantage of, while little visible light is actually used in making the image. This provides a strong infrared respose in the final image.
Part of the magic of infrared photos is how they portray a variety of subjects. In this graph, it is possible to see how much more IR light foliage (trees and grasses) reflects, compared to other elements (though many of the other elements reflect more IR light than available light). It also shows how much darker the ocean is in the IR range, predominantly because of the strong blue colour.
The most important factor in getting a good image with the infrared modified digital SLR cameras is starting out with a good white balance (this is assuming there is a good exposure as well). Outdoors, this is easiest to create off of foliage (which the IR camera renders as near white when correctly balanced). Indoors, with both available light or studio flash, I set the white balance off something like white sheets. This increases the brightness of the blue and green channels (see the histogram examples below), resulting in a more neutral image. If the white balance is inaccurate, it can always be reset in RAW processing.
White Balance Setting in RAW Conversion
With a conventional colour digital camera, using auto white balance (5500k, 0 tint) during RAW conversion, the resulting image has good colour balance and a broad range of tones between black and white, with the majority of the pixels being in the middle highlights (the blue, green and red spikes). Generally, this is the goal for basic RAW processing (good colour, with a detailed black and white in the image).
This result, shown here, is the combination of a good camera exposure, and careful RAW processing, which leads to the final image having highlights with fine, bright tones, and the shadows having deep, dark details.
With the infrared converted camera using an auto white balance (5500k, 0 tint) during RAW conversion, the image has a great degree of separation between the three colours that form the image (red, green and blue). This leads to an extreme imbalance in the image colour, overexposure in the red channel, and underexposure in the blue and green. Without correcting, the resulting image has an extreme red cast, and if converted to black and white would either have chalky whites (if we used a red conversion) or dull and gray (if blue and green were used). This simulates what the image would have looked like if it had been made in full colour, but it is a false representation, as the camera is not sensitive to any visible light, due to the filter over the sensor.
With the infrared converted camera using a manual white balance (2000k, -71 tint) during RAW conversion, the three colour channels are more evenly distributed. in this case, the manual white balance was created based upon the model’s skin. This gives the image a more neutral colour balance and a more evenly distributed tonality, which helps create a more pleasing final image.
The resulting image has a more neutral overall colour, but lacks most of the dark tones. More often than not, digital infrared images have a more compressed (lacking in blacks) tonal range than a full colour image, and need histogram correction in RAW conversion to maximize the tonal range. With indoor images especially, infrared images tend to pack the tones in the highlights, with little or any tones below the middle of the histogram. It is only in image post-production that this tonal range can be extended into a full range.
The final image with the infrared converted camera uses a manual white balance (2000k, -71 tint) and is desaturated* in RAW file conversion.
A histogram adjustment maximizes the tonal range, and an curve gives the final adjustments to the tonal distribution. The end product is a 60mb 16-bit TIFF file that only requires dust-removal, resizing and sharpening before use.
*it should be noted that for conventional digital photography, desaturation is the weakest way to create a monochrome image, with either a channel mix or hue/saturation layers approach being highly preferable