Photographing 40 Oil Paintings Under Mixed Light
You are photographing 40 framed oil paintings for a printed catalogue. The camera is a 24-megapixel body on a tripod with a 50 mm lens, positioned 1.2 m from each canvas. The room has two LED panels: a 5600 K panel on the left and a 3200 K panel on the right, and there is a large window behind you that cannot be blacked out before the end of the day. On the first test shots the left third of each canvas looks blue, the right third looks orange, highlights on the varnished surface blow out into white patches, and when you raise ISO to 3200 to hand-hold at 1/60 s the shadow areas of the dark canvases show visible grain. The catalogue will be printed at 300 ppi with canvases reproduced up to 210 mm wide.
1. Explain what is causing the colour variation across a single canvas and what you would change at capture time.
Read the worked answer
Two sources of different correlated colour temperature are lighting the same surface, plus uncontrolled daylight. The camera applies one white balance to the whole frame, so it cannot neutralise a 5600 K and a 3200 K source simultaneously: whichever it is set for, the other side goes the opposite way. That is exactly the blue-left / orange-right split.
Capture-time fixes, in order of effectiveness:
- Remove the mixture. Gel the 3200 K panel to daylight so both panels match, or switch one off and use a single source plus a white reflector.
- Flag the window. Daylight is a third, drifting colour - it changes with the weather and the hour, so no single balance will hold.
- Shoot RAW and place a grey card in the first frame of each canvas, then copy that custom balance across the set.
- Fix any residual unevenness in post with a local/graduated correction, not one global slider.
Do not try to correct a split cast by painting colour on the file; that is salvage, not capture.
2. The varnish highlights are clipping and the shadows are grainy. Give your settings strategy and justify it.
Read the worked answer
These are two separate faults with two separate causes.
Blown highlights. If the specular is at 255 in all three channels, the data is gone and no slider recovers it. Prevent it:
- Fit a polarising filter - it kills most of the glare bouncing off varnish.
- Move the panels further away and more to the side; angle of incidence equals angle of reflection, so light arriving at 45° reflects away from the lens.
- Read the histogram on a tethered preview and expose so the brightest specular sits just below clipping.
Grain. ISO 3200 is the cause and it is unnecessary - the camera is on a tripod. Drop to base ISO (100–200), lengthen the shutter to roughly 1/8–1/2 s, and use a 2-second timer or remote release with mirror lock-up to avoid shake. Stop down to f/8 for edge sharpness.
If a canvas genuinely exceeds the sensor's dynamic range, bracket ±2 EV and merge rather than pushing one noisy file.
3. The client also wants the largest canvases used on a poster as a 300 mm-wide detail. Assess whether the capture resolution supports this and note any capture-side quality risks.
Read the worked answer
Target: 300 mm at 300 ppi = 300 ÷ 25.4 ≈ 11.81 in × 300 ≈ 3543 px wide.
A 24 MP sensor gives roughly 6000 × 4000 px, so at the catalogue size (210 mm = 2480 px) you have about 2.4× the linear resolution needed and room to crop or correct perspective. Even the poster detail at 3543 px is within the sensor's output if the canvas fills the frame - but only just, so crop sparingly.
Capture-side risks that matter more than megapixels:
- Keystone and barrel distortion. The 50 mm at 1.2 m will show both. Level the tripod, square the camera to the frame edges with a grid overlay, and correct in RAW, cropping afterwards.
- Focus. Autofocus can lock onto glass or varnish; focus manually at the canvas centre and confirm at 100% on woven texture, not on the frame.
- Diffraction and depth of field. f/8 is a reasonable balance; check the corners, not the centre.