Automatic Focusing Based on Digital Image Processing

With the rapid development of computer hardware and digital image technology, real-time image processing has become possible. The computer collects a series of digital images through the lens and CCD, processes each frame of image in real time, determines whether the focus is accurate, whether the imaging is clear, and gives feedback signal to control the operation of the lens, automatic Focusing is completed until the collected image meets the requirements.

Automatic Focusing based on image processing has the following features:Two advantages:

First, focus is more intelligent, and focus criteria are more flexible and diverse. The focus Detection Method Based on simulated images only uses the contrast between the tested object and the background (the gradient of the contour edge) as the criterion for determining whether the image is clear. Through digital image processing, not only gradient information can be used, but also various other valid information in the image can be extracted for judgment, such as frequency and phase. For images with high-frequency information, in general, the more accurate the focus, the higher the image signal frequency, the more acute the edge; when the focus is removed, the frequency is reduced, and the edge is relatively smooth. In addition, due to the flexibility of the computer to process images, different criteria can be selected for focusing based on different usage requirements. For example, sometimes the target is only a part of the image, rather than the degree of clarity of the entire image. In this case, the image must be processed and extracted based on the local contrast (edge gradient.

Second, the computer can be used to conveniently control the operating execution mechanism, thus avoiding complicated focusing circuits and mechanisms. The computer interface and bus technology are very mature. The control signals are given through the software to directly control the operation of the motor animal mirror, which is not only flexible and convenient, but the response speed meets the focusing requirements, it can also greatly simplify the circuit and motion mechanism.

One of the key problems for automatic focusing by Image ProcessingSelection of image definition evaluation functions. The focus evaluation function can describe the degree of focus of the optical system based on the numerical value of the image.Basic RequirementsYes:

(1) The evaluation function must be effective before auto focus can be completed quickly. Auto Focus should take less time than manual operations;

(2) evaluation functions should have the following characteristics: they have a peak value at a certain point and correspond to the optimum focus parameter, and monotonically decline when the object leaves the optimum point. If such features are available, a very effective search method can be used;

(3) When there is noise, it is better to evaluate the robustness of the function.

Focus rating functions mainly include the following three types::

(1) spectrum functions. The theoretical basis is that clear images contain more information than blurred images, so that people can better distinguish the details (Details mean that the images have clearly identifiable edges, in some cases, there are strong gray-level changes, and gray-level changes are more intense );

(2) entropy function. The entropy of an image with good focus is greater than that of an image without clear focus. Therefore, the entropy function can be used as a focus evaluation standard. That is, the greater the entropy, the clearer the image;

(3) gradient functions. In image processing, gradient functions are often used to extract edge information. Focus-good image, more sharp edge image, should have a larger gradient function value. Common gradient functions include variance operators, energy gradient operators, and Laplace operators.

In actual optical systems, a targeted experiment is required.