Autofocus technologyIt draws on the intelligence of biological vision.It mimics the human visual system’s ability to focus rapidly and is a biomimetic design.

• From the three elements of exposure to the three elements of focus 

  As previously mentioned, the three key elements of automatic exposure are aperture, shutter speed, and ISO. Similarly, in autofocus, depth of focus, focal length, and depth of field (DoF) are three closely related optical concepts that collectively influence the range of sharpness and the visual impact of an image.

• Ideally, when light rays parallel to the optical axis enter a convex lens, all the rays converge at the focal point and then diverge in a conical pattern.

• Before and beyond the focal point, light rays begin to converge and diverge, causing the image of a point to blur into an enlarged circle. This circle is known as the circle of confusion. When the diameter of the circle of confusion becomes so small that it is indiscernible to the human eye, it is referred to as the permissible circle of confusion. There is one permissible circle of confusion on either side of the focal plane, and the distance between these two circles defines the depth of field.Depth of focus refers to the range along the optical axis of the lens within which the image remains reasonably sharp.

• The side corresponding to the subject being photographed is where depth of field lies. Depth of field is divided into foreground depth of field and background depth of field. It varies with the lens’s focal length, aperture setting, and shooting distance. For a fixed focal length and shooting distance, the smaller the aperture, the greater the depth of field.

• Autofocus algorithm

  The human eye’s autofocus capability is primarily achieved through the coordinated action of the ciliary muscle and the lens. This process occurs without conscious control by the brain; it is instantaneous and imperceptible.

• Camera autofocus (AF) refers to a closed-loop control system in which the camera uses a sensor to detect the focus status and drives the lens motor to adjust the lens group position, thereby forming a sharp image of the subject on the imaging plane. It can be categorized into…Image Input,Image Focus Detection,Image Focus ControlSuch a closed-loop control process. Evaluating the strengths and weaknesses of an autofocus algorithm primarily…

  It lies in focusing accuracy and focusing speed.

• This paper will focus on image focus detection; note that the classification below does not follow the chronological order of applications.

Autofocus technologies can be broadly categorized into passive and active types: passive approaches rely on analyzing ambient light and include phase‑detect autofocus (PDAF) and contrast‑detect autofocus (CDAF); active approaches, on the other hand, depend on emitting probing signals, with common techniques such as infrared ranging, ultrasonic ranging, and laser autofocus.

Active focusing

    Lidar（LiDAR）Its autofocus technology constructs a high‑precision 3D point cloud of the target area by emitting and receiving laser pulses in real time. Based on point‑cloud density and depth information, the system drives the focusing mechanism to adjust the focal length within milliseconds, ensuring consistently sharp imaging even in low‑light or complex‑texture environments. Compared with traditional contrast‑detection autofocus, LiDAR‑based focusing is faster and more adaptable, and has been widely adopted in mobile imaging, drone‑borne mapping, autonomous driving, and other fields. Below is…DJI’s LiDAR.

Passive autofocus

   Phase-detection autofocus:Phase‑detect autofocus (PDAF) draws inspiration from how humans perceive the world with both eyes. Simply put, PDAF works by splitting light to form two images, A and B, and calculating the phase difference between them; its ranging principle is analogous to the triangulation method used in binocular disparity.

In 1977U.S. company Honeywell has filed a patent for TCL (Through-the-Camera-Lens) phase‑detection autofocus. This system employs a ceramic substrate on which 48 CCD photosensitive elements are mounted, arranged in 24 pairs. Each pair consists of two staggered groups, A and B, with each group receiving light from the left and right halves via an overhead compound‑eye lens. Different CCD configurations correspond to different aperture settings, and rapid focusing is achieved by comparing the signal differences between the A and B groups.

Phase-detection autofocus comes in two forms: TCL phase detection and lens‑separated phase detection. In cameras, this manifests as two distinct approaches: one uses a beamsplitter and an optical baffle to split light onto separate AF sensors; the other achieves light splitting by modifying the pixel structure of the image sensor.

In DSLR systems, phase-detection autofocus typically relies on a dedicated AF module housed within the camera body, which uses a mirror and a beam splitter to redirect a portion of the incoming light to a specialized sensor. In mirrorless systems, phase-detection pixels are integrated directly into the image sensor, enabling the camera to simultaneously capture images and assess focus.Although the two approaches may differ in their architecture, they are both, at their core, phase-detection autofocus technologies. To put it simply:

Phase-detection autofocus is the most widely used autofocus method in DSLR cameras. This focusing system typically comprises hardware such as a reflex mirror, microlenses, and multiple image sensors, with its focusing principle illustrated in the figure.

Simply put, the light beam passing through the lens is split into two parts, each forming an image on a separate set of imaging sensors. When focus is accurate, the two images are identical. If focus is off, the two images will be displaced; by comparing the two images to measure this displacement, the degree of focus error can be determined.

Traditional DSLRs use a separate AF module,The PDAF system has only dozens of focus points, and there are…The phase signal is weak in low light.…limitations such as weak edge‑based autofocus. It has also accelerated the technological shift from DSLRs to mirrorless cameras. The conventional DSLR’s dedicated AF module and PDAF system indeed suffers from technical constraints, including a limited number of focus points, weak phase‑detect signals in low light, and insufficient edge‑detection performance. These limitations have collectively driven the camera industry’s transition from DSLRs to a mirrorless (MILC) architecture.

Shield PD: In 2014, Sony proposed selecting several pairs of pixels on the CMOS surface and masking half of each pixel with a metal or black light‑blocking layer, thereby forming two sets of half‑sensitive pixels, designated L and R.

Super PDSamsung proposed in 2019 to use a single microlens to cover two pixels, enabling on-chip phase detection.

Dual PD: In 2016, Canon was the first to adopt this technology on its EOS 70D DSLR, splitting each pixel into left and right halves to enable simultaneous shooting and autofocus. It delivers fast, accurate performance in low light with minimal image quality loss, making it the most “flagship‑level” phase‑detect autofocus system currently available on smartphones.

It should be noted that, in the absence of noise, strong reflections, low contrast, and other interfering factors, phase‑detection (PD) can accurately compute the disparity between the left and right images and achieve fast autofocus by calibrating the relationship between PD values and lens displacement. However, in real‑world scenarios, PD measurements often contain errors. Consequently, when a PDAF system obtains measurement results, it must also assign a PD confidence score based on the current environmental conditions; downstream modules then proceed to…Reliability assessment for achieving precise focusing.