How to Clean Up Blurry and Shaky Photos on iPhone

Types of Blur: Motion, Focus, Low Light

Not all blurry photos are created equal. Understanding the three main types of blur helps you make better decisions about which photos to keep and which to delete. Each type has distinct visual characteristics, different causes, and different detection methods.

Motion Blur

Motion blur occurs when the camera or the subject moves during the exposure. On an iPhone, even tiny hand movements can introduce noticeable shake, especially in low light when the shutter stays open longer. The iPhone's sensor-shift optical image stabilization (available since iPhone 12 Pro) reduces this significantly, but it cannot eliminate it entirely. You can identify motion blur by its directional streaking pattern: edges smear along the path of movement. In bright daylight with fast shutter speeds above 1/500s, motion blur is rare. But indoors or at dusk, when the shutter drops to 1/30s or slower, even a heartbeat can blur a shot. Studies suggest that roughly 15 to 20 percent of casual handheld photos suffer from some degree of motion blur.

Focus Blur

Focus blur happens when the lens locks onto the wrong distance. Unlike motion blur, it is uniform across the image rather than directional. The autofocus system on modern iPhones is fast, but it can still misfire when there are multiple subjects at different depths, when shooting through glass, or when the subject has low contrast (like a white wall or a foggy sky). Portrait mode deliberately introduces focus blur as bokeh in the background, which is why a naive blur detector that simply measures softness will flag portrait shots as false positives. A good blurry photo cleaner must distinguish intentional artistic softness from genuine focus errors. You can spot focus blur by checking whether a different part of the image is sharp while your intended subject is soft.

Low-Light Blur

Low-light blur is a combination problem. In dim environments, the camera extends the exposure time, which amplifies motion blur. Simultaneously, the sensor raises its ISO, which introduces noise that obscures fine detail and can be mistaken for softness. Night mode on iPhone mitigates this by capturing multiple frames and merging them computationally, but older photos taken before Night mode was available, or photos taken with apps that bypass it, often show this dual degradation. These images are grainy, have reduced contrast, and display a characteristic softness that blends noise artifacts with genuine blur. For automated detection tools, low-light shots are among the most challenging to evaluate because the noise can mask or exaggerate the underlying blur.

How Vision Framework Detects Blur

Apple's Vision framework is a suite of on-device computer vision algorithms that can analyze images without sending data to any server. While Vision does not have a single "detect blur" API, it provides several building blocks that, when combined, produce reliable blur detection. This is the foundation that TidyByte uses as a blurry photo cleaner.

Laplacian Variance

The most widely used technique for measuring image sharpness is Laplacian variance. The Laplacian operator is a mathematical filter that highlights areas of rapid intensity change, which is exactly what sharp edges produce. When you apply it to an image, sharp photos produce a high-variance output (strong, well-defined edges) while blurry photos produce a low-variance output (weak, diffuse edges). This method is particularly good at detecting focus blur because it evaluates the entire image at once. A Laplacian variance below approximately 100 on a 1-megapixel input typically indicates a blurry image, though the exact threshold depends on the scene content.

Optical Flow Analysis

Vision's VNOpticalFlowRequest computes the apparent motion of pixels between regions of a single image. While optical flow is traditionally used for video, applying it within a single frame can reveal directional blur patterns that are characteristic of motion blur. If the flow vectors are strong and aligned in one direction, it strongly suggests camera shake. This signal is especially useful for distinguishing motion blur from focus blur, which produces no directional flow pattern.

Feature Print Generation

VNFeaturePrintRequest generates a compact mathematical representation of an image's visual content. While it was designed for similarity comparison rather than blur detection, the texture information embedded in the feature print correlates with image quality. Blurry images produce feature prints with less distinct texture patterns than sharp ones. By analyzing the entropy and distribution of feature print values, you can extract a quality signal that complements the other methods.

Edge Density via Contour Detection

VNDetectContoursRequest identifies the boundaries of objects in an image. Sharp photos produce many well-defined contours with high contrast along edges, while blurry photos produce fewer, weaker contours. By measuring the density and strength of detected contours relative to the image size, you get another independent signal. This approach is less sensitive to noise than Laplacian variance, making it useful for evaluating low-light images where noise might otherwise skew the results.

TidyByte's Multi-Factor Analysis

TidyByte combines the signals above into a single multi-factor blur score. Rather than relying on any one technique, it runs four independent analyses in parallel and merges the results into a composite score from 0.0 (sharp) to 1.0 (severely blurry). This approach eliminates the weaknesses of individual methods and produces a robust blurry photo cleaner that works across all types of blur.

The entire analysis runs on-device using Apple's Vision framework and Core Image, so your photos never leave your iPhone. TidyByte processes images through its VisionAnalysisService, which operates as a Swift actor to keep the main thread responsive. The service accepts CGImage inputs, which means it does not require photo library access just to perform the analysis. Each photo is scored in under a second on modern iPhone hardware, and the service yields to the Task scheduler every 20 photos to keep the UI responsive during batch scans. You can also adjust the blur sensitivity threshold in settings to control how strict the detection is. A threshold around 0.65 works well for most users, but you can lower it to catch only the worst offenders or raise it to flag slightly soft images as well.

Reviewing Detected Photos

Once TidyByte has scanned your library, it presents the results in its signature swipe interface. Each detected photo shows its blur score and the primary blur type (motion, focus, or low-light), so you can make an informed decision. The swipe-left-to-delete, swipe-right-to-keep interaction makes reviewing fast: most users can process 100 photos in under three minutes.

For photos you are unsure about, tap to view them full-screen before deciding. The review screen also shows the photo's metadata, including the capture date, camera settings, and file size, which can help you judge whether a slightly blurry shot is worth keeping for sentimental reasons. TidyByte does not auto-delete anything; every decision is yours.

A practical tip: start with the highest blur scores first. Photos scored above 0.85 are almost certainly unusable, so clearing those out quickly reduces the pile and builds momentum. Then work through the moderate range 0.65 to 0.85, where some photos may be borderline. TidyByte also offers a Select All option for each blur category, which is useful if you want to bulk-delete a large batch of clearly blurry shots and only review the ambiguous ones individually. Combined with the app's duplicate detection feature, you can often reclaim significant storage in a single session.

Edge Cases and False Positives

No automated detection system is perfect. There are situations where blur analysis can produce unexpected results, and understanding these edge cases helps you use the tool more effectively.

• Portrait mode and bokeh. Intentional background blur in portrait photos can trigger false positives. TidyByte mitigates this by checking whether the blur is uniform across the frame or localized to the background. If the subject's face and eyes are sharp, the image is likely a portrait rather than a mistake.
• Intentional motion blur. Long-exposure artistic shots, light trails, and waterfall smoothing are deliberate creative choices. These often have strong directional flow that mimics camera shake. Because the blur is typically uniform and aesthetically composed, TidyByte's multi-factor approach usually assigns a lower score, but some may still appear in results.
• Screenshots and graphics. Screenshots, memes, and text-heavy images have different frequency characteristics than photographs. TidyByte's analysis is calibrated for photographic content, so screenshots generally pass through without being flagged, even if they contain small text that appears soft at full resolution.
• Extreme crops. Heavy digital zoom or aggressive cropping amplifies any existing blur and introduces pixelation, which some blur metrics can mistake for sharpness. TidyByte handles this by combining the edge density signal with the Laplacian variance, which helps distinguish between genuine sharpness and blocky artifacts from upscaling.
• Panning shots. Photos where you deliberately track a moving subject (like a car or cyclist) have a sharp subject with a motion-blurred background. These can confuse detectors that evaluate the whole frame. TidyByte's feature print analysis helps here because it captures the texture contrast between the sharp subject and the blurred background, typically scoring these lower than true failed shots.

If you encounter false positives, you can adjust the blur sensitivity in TidyByte's settings. You can also look into the app's duplicate detection and screenshot cleanup features for a more comprehensive approach to library maintenance.

Maintaining a Clean Library

Cleaning blurry photos is not a one-time task. New blurry shots accumulate every time you take photos in challenging conditions: at concerts, on moving vehicles, in dimly lit restaurants, or simply when your hands are cold. Building a regular maintenance habit keeps your library manageable and your storage free.

A good cadence is to run a scan once a month. On a library of 10,000 photos, TidyByte's blur analysis typically completes in a few minutes because the Vision framework operations are highly optimized for Apple silicon. The app also caches its analysis results using its SwiftData models, so re-scanning after adding a few hundred new photos only processes the additions, not your entire library.

Beyond automated detection, you can reduce future blur at the source. Enable Live Photos, which capture 1.5 seconds of motion around each shot and let you pick a sharper frame later. Use burst mode for action shots and select the best frame. In low light, hold the phone steady against a surface or use the timer to avoid shutter-press shake. And keep your lens clean; a fingerprint smear softens every photo you take.

According to industry estimates, the average smartphone user has over 2,000 photos on their device, and studies suggest that 10 to 15 percent of those are blurry, duplicates, or otherwise unwanted. That means roughly 200 to 300 photos sitting in your library that serve no purpose other than consuming storage and making it harder to find the shots that matter. A few minutes with a blurry photo cleaner each month can recover that space and keep your gallery organized.

Frequently Asked Questions