By Nila Kartika Wati 
Capturing iconic urban landscapes and architectural marvels free from the constant ebb and flow of human traffic presents a significant challenge for photographers. The pursuit of an empty plaza, a deserted street, or a serene monument often clashes with the reality of bustling public spaces. To overcome this, photographers employ sophisticated techniques, primarily long exposure photography with neutral density (ND) filters and post-processing median stacking, to effectively "erase" transient elements like pedestrians from their frames, transforming vibrant scenes into timeless, solitary visions.
The Elusive Empty Frame: A Photographer’s Quest
The fundamental problem lies in the photographic sensor’s function: it accumulates light over the duration of an exposure. A person moving across a scene will only occupy a specific pixel for a fleeting moment. For instance, imagine a 35mm lens on a full-frame camera, positioned approximately 20 meters from a plaza. Each pixel in the resulting image file will correspond to roughly 3.4 millimeters of ground space at that distance. A person walking at an ordinary pace, clearing their own shoulder width (approximately 0.45 meters or 130 pixels), will occupy any single pixel for about 0.32 seconds. Across a four-minute (240-second) exposure, this translates to a mere 0.13 percent of the total exposure time for that pixel. This minute fraction is the core principle behind the disappearance act. The ability to calculate and understand this "dwell time" for individual and aggregated movements allows photographers to predict which elements will vanish and which might persist, potentially compromising the desired aesthetic.
The Physics of Disappearance: Dwell Time and Pixel Occupancy
A sensor pixel functions as a photon counter, accumulating light without memory of arrival time. Therefore, if an object covers a pixel for a duration ‘t’ within a total exposure ‘T’, it contributes approximately t/T of that pixel’s final value. This contribution is also weighted by the object’s luminance and color relative to the background. Two primary factors dictate the outcome: the fraction of exposure time an object covers a specific spot, and its color and luminance contrast against the background. A dark coat against bright pavement provides the most pronounced signal difference. However, color, often overlooked, plays a crucial role; a red coat on gray stone might meter similarly in overall brightness but will leave a distinct colored stain because the camera records red, green, and blue channels separately, and these differences persist. The size of an object only matters inasmuch as it determines how many pixels it touches and the time it takes to clear a single one.
Considering the plaza example: A 35mm lens on a full-frame sensor covers roughly 20.6 meters horizontally at a 20-meter distance. A 6,000-pixel-wide image file thus divides this into approximately 3.4 mm per pixel. A person, about 0.45 meters wide, occupies around 130 pixels. At a standard walking pace of 1.4 m/s, they clear their width in 0.32 seconds. In a 240-second exposure, this is 0.13 percent of the total time. A perfectly black silhouette blocking a pixel for this brief interval removes 0.13 percent of the light. In a well-exposed midtone, the pixel’s inherent shot noise often exceeds this value several times over, effectively burying the single walker’s contribution. Consequently, a lone pedestrian typically becomes imperceptible in an ordinary print or screen view. However, this disappearance isn’t absolute; a moving person leaves a faint, smeared band along their path, which, despite being below the per-pixel noise floor, can sometimes be recovered with aggressive post-processing like local contrast enhancements or denoisers. The robust claim is that a single pass is generally imperceptible under normal viewing, not that it leaves no trace in the raw data.
Method One: The Enduring Power of Long Exposure Photography
The classic technique to achieve an empty scene involves a single, very long exposure, often requiring powerful neutral density (ND) filters in bright conditions. ND filters reduce the amount of light entering the lens without altering color, effectively allowing for significantly longer shutter speeds. Each "stop" of an ND filter halves the light, meaning the exposure time multiplier increases exponentially. For instance, a 10-stop ND filter multiplies the required exposure time by 1,024, while a 15-stop filter extends it by an astonishing 32,768.
These filter strengths lead to dramatically different photographic sessions. On a sunny midday street, metering at 1/200 s at f/11 and ISO 100 (per the sunny 16 rule), a 10-stop filter would yield an exposure of only 5.1 seconds – insufficient to erase most movement. To achieve truly useful erasure, closer to three minutes, a 15-stop filter would be necessary, extending the exposure to approximately 164 seconds. A 10-stop filter only reaches a four-minute exposure when the ambient light is already very low, such as deep shade or during blue hour (around EV 9, or 1/4 s at f/11). Heavy overcast conditions, typically around EV 12 (1/30 s at f/11), would only extend a 10-stop exposure to about 30 seconds, far short of the multi-minute durations often required. Understanding these calculations beforehand prevents photographers from encountering frustratingly smeared images after investing significant time.
Leading manufacturers like Lee Filters offer a range of professional-grade ND filters, including the Little Stopper (6 stops), Big Stopper (10 stops), and Super Stopper (15 stops). Screw-on alternatives from brands like Breakthrough Photography (X4 ND 3.0), B+W (Master 810 ND 3.0), NiSi (ND1000), and Haida (NanoPro ND1000) also provide 10-stop options. It’s crucial to verify the filter’s density number (e.g., 3.0 or 1000x for 10 stops) rather than relying solely on marketing names, as nomenclature can vary.
However, the long exposure method presents practical challenges. Focusing becomes difficult through extremely dark filters, often necessitating pre-focusing and switching to manual mode. Strong ND filters, particularly cheaper ones, can introduce a color cast, typically a magenta or blue tint, requiring correction in post-processing. Light leaks, especially with square filter systems, can fog the image. Furthermore, any non-transient motion, such as flags fluttering, leaves rustling, or fountain spray, will blur into a uniform, often unappealing, mush, losing all definition. While this "silky water" effect is often desired, it can be detrimental to other elements.
Method Two: Computational Eradication with Median Stacking
A more contemporary and often preferred approach among many working photographers for achieving empty public spaces is median stacking. This technique involves capturing multiple frames (typically 15 to 30) of the same locked-off composition at normal shutter speeds, spaced a few seconds apart, and then using software to calculate the per-pixel median value across the entire stack.
The power of median stacking lies in its statistical operation. For any given pixel, if 19 frames show the background pavement and one frame shows a person’s jacket, sorting these values will place the jacket’s outlier value at one end of the list. The median, being the middle value, will then select one of the pavement values, effectively discarding the outlier. This differs fundamentally from averaging (mean stacking), which would pull the pixel’s value slightly towards the outlier, potentially leaving a faint smudge. Median stacking is ideal for outlier rejection, while mean stacking is better for reducing random noise.
The number of frames is critical. For an odd-numbered stack, a clear majority ensures clean removal. For example, in a stack of 21 frames, if a person appears in 10 or fewer, the background will always constitute the majority, and the median will render the background. With an even-numbered stack, such as 20 frames, if a subject occupies exactly 10 frames, the median (calculated as the average of the two middle values) might result in a half-strength blend of the person and the background. Therefore, using an odd number of frames (e.g., 21 instead of 20) is often recommended to guarantee a clear majority for clean removal.
The workflow in Adobe Photoshop is straightforward: File > Scripts > Load Files into Stack, add the frames, and select Create Smart Object after Loading Layers. Optionally, Attempt to Automatically Align Source Images can correct minor tripod shifts, though this involves resampling and slight cropping. The crucial step is applying the blend mode: Layer > Smart Objects > Stack Mode > Median. Adobe’s documentation explicitly lists Median for both noise reduction and object removal, highlighting its dual utility. The software then processes the stack, rendering the scene devoid of transient crowds.
Median stacking offers several significant advantages over long exposure with ND filters. No physical filter is required, eliminating color casts, the difficulty of focusing in darkness, and light leaks. Shutter speeds remain fast enough to freeze desired motion, keeping elements like flags, leaves, and fountain spray crisp rather than blurring them. Changing light conditions are less problematic, as each frame is exposed normally, and a passing cloud shadow affecting only a minority of frames will be rejected by the median. Sensor heat, a concern with multi-minute exposures, is also less of an issue, though continuous shooting can still generate heat. Crucially, all original frames are preserved, providing flexibility for later adjustments.
The "cadence"—the interval between shots—is vital. Short gaps can be problematic; a pedestrian strolling at 1.4 m/s might only move a few meters in two seconds, potentially occupying the same patch of ground across multiple consecutive frames. A five-second gap, however, moves a pedestrian approximately 7 meters, usually enough to place them in an entirely different part of the frame. A session of 20 frames at five-second intervals, lasting about 100 seconds, is a reasonable default for moderately busy squares. Busier or slower-moving scenes may require longer gaps, more frames, or both, ensuring the clean background holds a majority in every pixel across the stack.
Strategic Considerations for Flawless Results
Achieving optimal results with either method demands meticulous attention to camera settings and environmental factors. Regardless of the chosen technique, a locked-down tripod is paramount. Any movement between frames, even a slight nudge, undermines the premise that each pixel is observing the same spot. While auto-alignment features in software can mitigate this, they introduce resampling, potentially reducing fine detail and cropping edges. Using a remote shutter release or the camera’s self-timer is essential to avoid camera shake.
Shooting at base ISO (the camera’s native, lowest sensitivity) is recommended. Many cameras offer "extended low" ISO settings (e.g., ISO 50), but these are often "pulled" ISO 100, meaning the sensor is overexposed and then digitally darkened, sacrificing highlight headroom—an undesirable trade-off when chasing longer exposures or clean data.
Aperture selection should prioritize sharpness, typically f/8 to f/11. Resisting the temptation to stop down further (e.g., to f/22) for increased exposure time is crucial, as diffraction at smaller apertures significantly degrades image sharpness, spreading light over a larger area and reducing effective resolution. If more light reduction is needed, additional ND filtration or waiting for lower ambient light is preferable.
Most modern mirrorless cameras and many DSLRs (e.g., Nikon D7500/D850, Canon EOS 5D Mark IV) include built-in interval timers, simplifying the process of capturing timed sequences. For cameras lacking this feature, an external timer remote, such as the Vello ShutterBoss II, provides programmable control over frame count and interval.
Lighting conditions greatly influence success. Overcast days and blue hour are often ideal. Overcast conditions provide consistent, diffuse light, preventing drastic exposure changes between frames in a stack or harsh shadows in a long exposure. Blue hour offers naturally lower ambient light, making it easier to achieve multi-minute exposures with a 10-stop ND filter and naturally reducing crowd density. Conversely, harsh midday sun is the most challenging, introducing rapidly moving shadows, extreme brightness ranges, and peak crowd numbers. Professional cityscape photographers often emphasize choosing the right time of day as a fundamental skill.
The Unyielding Subject: Addressing Persistent Presence
The most significant hurdle for both methods is the "immovable object"—a person or group who remains stationary within the frame for an extended period. For a single long exposure, someone posing for a photo, a street vendor, or a couple conversing for a significant portion of the exposure will not vanish; they will appear as a translucent ghost, sharp enough to be recognizable but soft enough to look like an error. If they remain for the entire exposure, they will appear solid, as if photographed with a fast shutter speed. The only recourse is to repeat the exposure, which is time-consuming and offers no guarantee.
Median stacking offers more robust solutions. If a subject occupies a pixel in more than half the frames (e.g., 12 out of 20), they will persist. The first solution is to extend the shooting session; if the subject eventually moves, adding more frames can restore the background’s majority. The second, and often more practical, solution leverages the preserved original frames. If even one frame in the stack shows the spot clear, that patch of clean background can be meticulously cloned or masked into the final median result. Since the camera remained static, the geometry aligns perfectly, eliminating the guesswork associated with traditional cloning.
Beyond people, other moving elements behave differently with each technique. A single long exposure typically streaks clouds into directional blurs and transforms water into a smooth, silky surface—effects often sought after. Median stacking, however, can yield stranger results. Moving water often smooths out as its highlights and shadows are rejected, but static textures can persist, and broken reflections might leave irregular patches. Drifting clouds in a median stack result in a per-pixel composite that may not correspond to any actual cloud shape, creating an often artificial appearance. For photographers desiring both silky water and an empty foreground, a hybrid approach—a long exposure for water and sky, and a median stack for the ground—blended with a mask along the horizon, is often the most effective.
Broader Implications and The Future of Imaging
The mastery of these techniques transcends mere technical proficiency, carrying broader implications for commercial, artistic, and even ethical considerations in photography. Commercially, they enable architectural firms, real estate developers, and tourism boards to present pristine, uncluttered views of their projects or destinations, conveying a sense of serenity and grandeur often impossible in reality. Artistically, they allow photographers to create timeless, contemplative images that emphasize structure and form over the fleeting human element, evoking solitude or a sense of history.
Ethically, the ability to "erase" people from public spaces raises questions about representation and authenticity. While generally accepted in artistic and commercial contexts focused on architecture or landscape, its use in photojournalism would be highly problematic, bordering on fabrication. This highlights the importance of context and intent in photographic manipulation.
As computational photography continues to evolve, we can expect even more sophisticated methods for separating transient elements from static scenes. Artificial intelligence and machine learning algorithms are already being explored to enhance object removal, potentially offering more intelligent ways to handle complex movements, retain desired elements, and even reconstruct hidden backgrounds with greater fidelity. These advancements promise to further empower photographers in their quest for the perfect, unpeopled frame.
Ultimately, both long exposure and median stacking are powerful tools, each with its strengths and weaknesses. The truly skilled photographer understands not just how to use them, but when. It’s a matter of thinking in fractions of occupancy rather than just shutter speeds. By assessing the traffic patterns, the slowest-moving elements, and the desired aesthetic, photographers can choose the optimal method or combination, transforming bustling public spaces into moments of pristine, timeless beauty.
