Image processing

In image processing is understood in computer science and electrical engineering, the processing of signals that represent images, such as photographs or frames of videos. The result of image processing can in turn be an image or a set of features of the input image (see image recognition ). In most cases images are viewed as a two dimensional signal, so that conventional methods can be used in the signal processing.

Image processing is to be distinguished from image processing that deals with the manipulation of images for subsequent display.

  • 5.1 Image Enhancement
  • 5.2 Object detection and tracking

Application examples of image processing

Image processing is now used in almost all fields of science and engineering disciplines, such as in modern microscopy, medical diagnostics, astronomy, mechanical engineering and remote sensing ( environmental monitoring, espionage). With methods of image processing objects are in the machine counted, measured, inspected objects or read encoded information. X-ray and ultrasound units can provide the image processing of images that can easily suggest the doctor. X-ray equipment in security zones to examine luggage and clothes automatically for dangerous items ( weapons, etc.). Another field is the quality assurance in manufacturing and production processes. The so-called in robotics bin picking is also supported by image processing.

Inspection and measurement of objects

Measuring chip in the semiconductor manufacturing: it is the position of the corner of a chip, for example, measured in an image. With this information, the die may be accurately positioned for assembly. In the chip - final errors are detected during soldering and bonding by a chip is moved in front of the camera and is compared to a golden sample that was previously learned by the image processing system. Or is it looking for clearly defined geometric patterns (circle, corner) for cracks or outbreaks.

Image processing in a beverage bottling plant: To check whether was introduced into each bottle all the same, is a picture of the bottle neck made ​​and measured the liquid edge. Before bottling, it is checked whether the bottleneck is cracked or chipped.

Measurement of adhesive in micromechanics: In the manufacture of camera modules for mobile telephones adhesive is applied to the lens holder. In order to ensure a uniform quality of production is checked in an image the shape of this adhesive. If the adhesive is not within certain tolerances, the device is discarded.

Reading coded information

Image processing can be read automatically encoded information from images. For example, can be read by OCR or information are extracted as plain text encoded in a Data Matrix form text. These functions are also used in the letter and parcel identification.

In the automotive manufacturing serial numbers of components in the data matrix are encoded form. If a module reaches a production area, is taking a picture of the code with a camera and read the code. This serial number provides machine of the production area of a server information such as the assembly is to be treated.

Objects in the image processing

Image processing methods generally expect image data as input. This image data can also be distinguished in their coding both in the nature of their creation, as. The type of development describes technical principle on which the image was formed. The most common here are reflection images as they arise at a camera or ultrasound. In addition, the projection images, such as X-rays, as well as Schematic images, such as maps and documents exist. When encoding bitmapped is the most common form in which the image data is represented by a two-dimensional raster of pixels. Another form are vector graphics that do not consist of a grid, but instructions included, like a picture of geometric primitives is to produce.

Differentiation from related fields

Related areas of image processing, the image processing, the Machine Vision and Computer Graphics. With image editing a more abstract view is placed on the change of images, while the image processing this is provided by the mathematical and algorithmic foundations, which are then used in the implementation of graphics software for image processing. This provides the image processing for machine vision. While the image processing of image data, in turn, generates image data or simple information, the Machine generated Look Image Data Image descriptions. The computer graphics in turn generated from image descriptions image data.

Operations of the image processing

In the operations for image processing is first necessary to distinguish between procedures that create a new image, and those that provide information about the image. The methods produce a new image can be distinguished by the size of the region of the input data. In addition, to distinguish whether the procedure obtains the basic structure of the image or changing them.

A common method to generate information from an image, the calculation of the histogram, which gives information about the statistical distribution of brightness in the image. Such a histogram can for example serve as a configuration for further image processing steps, or as information for a human user of software. More predictable information of an image, for example its entropy or average brightness.

Procedures that create a new image, can be distinguished because of their input data into point operations, neighborhood operations and global operations. The point operations use the color or brightness information at a given point of the image as input, calculate a new brightness value as the result and store it at the same point in the target image. Typical applications of point operations are for example the correction of contrast and brightness, color correction by turning the color space or the use of different thresholding. A point operation can be either homogeneous, which means that the coordinate of the source data is not included in the calculation, or they may be inhomogeneous, which for example allows an adaptive Levels. Neighborhood operations using both a point and a certain amount of its neighbors as input, calculate from them the result and write it to the coordinate of the reference point in the target image. A very common type of neighborhood operations are the convolution filter. Here, the brightness or color values ​​according to a filter kernel to be offset against one another in order to form the result. With this method, for example blur filter like the median filter, Gaussian filter or the binomial filter can be realized. Aliasing filter as edges of an image can be highlighted using derivative filters or Laplacian filter. The neighborhood operations are not limited to the convolution filter. By complex algorithmic treatment of the reference point and its neighbors, for example, other methods for smoothing such as filtering media, or for edge detection of the Extremalspannenfilter or the Prewitt operator can be realized. From morphological operators such as erosion and dilation operations, the opening, closing and therefore a Morphological smoothing can be defined. While smoothing over relatively simple neighborhood operations can be realized, is a deconvolution and sharpening the image a more complex task. Neither point still neighborhood operations modify an image in its size or its basic structure. This is achieved by geometric pictures operations such as scaling, rotation or translation of an image, Anisotropic filtering is required here, and the interpolation is a key criterion for image quality. The geometric images operations are part of the global image operations which use the complete image as input data. Another representative of the global image operations is the Fourier transform, the image in the frequency domain is converted, in the application of linear filter means a lower cost.

The time requirement of the above operations is very much dependent on the image resolution.


Image enhancement

Image with artificially added Salt -and- Pepper noise.

Application of a 3 × 3 Gaussian filter to the noisy image.

Application of a 3 × 3 median filter to the noisy image.

Object detection and tracking

Pictures of Image processing