Although millions of pixels can only look at the picture and sigh, there is nothing to do; Another example is the blocking effect caused by high compression ratio coding in order to save bandwidth, resulting in the decline of image quality.
How to solve the sharp decline in image quality caused by these extreme environments and see more clearly in these harsh environments is an urgent problem to be solved by video image technology. In particular, video surveillance is more and more widely used in life. People pay special attention to the application and development of video processing technology under the new situation.
Digital video and digital images have higher resolution than traditional images and video, convenient processing, easy operation and sorting. However, due to the insufficient performance of some equipment, objective conditions and other factors, in the actual video monitoring application, there will still be some problems, such as blurred video images and unable to capture key information.
In the process of video image processing, some negative effects are brought to the application of video image processing technology due to operation technical problems or objective factors, which reduces the level and quality of processing technology.
Four technologies of video image processing technology
The process of video image processing involves the collection, transmission, processing, display and playback of video image data. These processes together form an overall cycle of the system and can operate continuously. In the range of video image processing technology, the most important ones include image compression technology and video image processing technology.
At present, the mainstream video image processing technologies in the market include: intelligent analysis and processing, video fog penetration and antireflection technology, wide dynamic processing and super-resolution processing. The above four processing technologies are introduced below.
Intelligent analysis and processing technology
Intelligent video analysis technology is an important means to solve the technical problems of big data screening and retrieval in the field of video surveillance. At present, domestic intelligent analysis technology can be divided into two categories: one is to detect the movement of objects in the picture by foreground extraction and other methods, and distinguish different behaviors by setting rules, such as line mixing, object legacy, perimeter, etc; The other is to use pattern recognition technology to model the objects to be monitored in the picture, so as to detect specific objects in the video and related applications, such as vehicle detection, people flow statistics, face detection and so on.
Video fog penetration and antireflection Technology
Video fog penetration and antireflection technology generally refers to making the blurred image caused by fog, moisture and dust clear, emphasizing some interesting features in the image and suppressing the uninterested features, so as to improve the quality of the image and enrich the amount of information. Due to the bad conditions such as fog and haze, rain and snow, strong light and dark light, the image contrast of the video monitoring image is poor, the resolution is low, the image is blurred, and the features cannot be recognized. The image after antireflection processing can provide good conditions for the next application of the image.
Dynamic algorithm of digital image width
Wide dynamic range is a basic feature in digital image processing, which occupies an important position in image and visual restoration, and is related to the imaging quality of the final image. The dynamic range is mainly determined by the protection semaphore and the average noise ratio, and the dynamic range can be defined from the perspective of light energy.
Digital signal processing will be affected by exposure effect, illumination and intensity in exposure. The dynamic range is closely related to the depth of the pattern. If the image dynamic range is wide, the brightness change is obvious during image processing, but if the dynamic range is narrow, the change of brightness and darkness is not obvious during brightness conversion. At present, the wide dynamic range of images is widely used in video surveillance, medical imaging and other fields.
Super resolution reconstruction technology
The most direct way to improve the image resolution is to improve the sensor density of the acquisition equipment. However, the price of high-density image sensor is relatively expensive, which is difficult to bear in general applications; On the other hand, the imaging system is close to the limit due to the limitation of its sensor array density.
The effective way to solve this problem is to improve the spatial resolution of the image by using the software method based on signal processing, that is, Sr (super resolution) image reconstruction. Its core idea is to use the time bandwidth (obtain multiple image sequences of the same scene) to change the spatial resolution, so as to realize the conversion from time resolution to spatial resolution, The visual effect of the reconstructed image is more than any low resolution image.
Two techniques of image enhancement
In addition to video image processing technology, image enhancement technology can purposely emphasize the overall or local characteristics of the image for the application of a given image, make the original unclear image clear or emphasize some interesting features, expand the difference between the features of different objects in the image and suppress the uninteresting features. In the development of image enhancement technology, suppressing or eliminating image noise points occupies a very important position, and many special "denoising" algorithms have been developed.
Firstly, the spatial domain image enhancement method is introduced. Spatial domain enhancement refers to the process of enhancing the pixels constituting the video image and directly operating on these pixels. There are mainly the following methods:
Basic grayscale transformation: mapping image pixel values from one range to another, including linear transformation, logarithmic transformation and power transformation. For example, gamma correction is a power transformation. Through gray-scale transformation, the gray-scale difference between different pixels can be improved, the contrast can be improved, and it is more conducive for human eyes to recognize the details. At the same time, this method is also the basis of some other advanced methods.
Histogram processing: the method of obtaining a new image with uniform or specified gray histogram through some transformation of the original image. It is one of the most commonly used and important algorithms in image enhancement algorithms. Based on the probability theory, it uses gray point operation to realize the transformation of histogram, so as to achieve the purpose of image enhancement. Histogram homogenization can effectively improve the dynamic range of the image, improve the contrast, and is more conducive to human eye recognition of details.
Smoothing spatial filtering: smoothing spatial filtering is mainly used to filter image noise and smooth image. There are many smoothing filtering methods, such as linear smoothing filtering, including mean filtering, nonlinear smoothing filtering, such as median filtering. Linear filtering has a good smoothing effect. It can filter noise, but it will also blur the edge details.
Nonlinear filtering is an improvement of linear filtering. It will adopt different strategies according to the state of pixels, which can eliminate some isolated noise, has little impact on image details, but will bring some distortion to the edge of the image. In order to overcome the shortcomings of the above two algorithms, people put forward many improvement schemes, introduce adaptive smoothing algorithm, and take into account noise filtering and image detail maintenance through various methods.
Sharpening spatial filtering: in contrast to smoothing spatial filtering, sharpening spatial filtering is to highlight the details in the image or enhance the blurred details. Sharpening spatial filtering is mainly realized by first-order and second-order sharpening filters, such as gradient method, Laplace operator filtering and so on. Sharpening spatial filtering will improve the image details, but it will also enlarge the noise points.
Secondly, frequency domain image enhancement method. Frequency domain image enhancement is to treat the image as a two-dimensional signal and transform it to frequency domain for filtering and enhancement. There are many methods to transform images from spatial domain to transform domain, such as Fourier transform, Walsh Hadamard transform, cosine transform, K-L transform and wavelet transform. Fourier transform and wavelet transform are common transform methods for image denoising.
Low pass filtering: similar to spatial smoothing filtering, the high-frequency part is filtered to achieve the purpose of denoising. Commonly used are Butterworth low-pass filter, Gaussian low-pass filter and so on.
High pass filtering: similar to sharpening spatial filtering, it retains more high-frequency parts to improve image details, but it will also enlarge noise.
The application of frequency domain image enhancement in the field of surveillance video is limited due to its high computational complexity due to frequency transformation. Many other image enhancement and denoising methods have good performance in noise point removal, edge detail processing and contrast enhancement.
However, due to its high computational complexity or strong pertinence, it is not suitable for use in the field of video surveillance, but more for special image processing systems, such as medical images, remote sensing and other fields. However, with the continuous development and progress of hardware equipment and image technology, more new video enhancement technologies will be applied to video surveillance.
Source: China Public Security