Color analysis of image processing

The CIE (Commission Internationale de l`Eclairage) is an abbreviation of the International Lighting Association. It sets an international standard for measuring color and measures color values. The CIE developed L*, a*, and b* values ​​to measure color values. This measurement method is called CIELAB. L* stands for lightness, from bright (L*=100 at this time) to darkness (L*=0 at this time) ) Changes between. The A* value indicates the change in color from green (-a*) to red (+a*), and the b* value indicates the change in color from yellow (+b*) to blue (-b*). After using the system, any color can find a corresponding position on its chart, see picture. With the change of L*, a*, b* and E*, the angle L*, angle a*, angle b* and angle E* change with the angle E*=angle (angle L*2 + angle a* 2+ angle b*2). This value represents the color value of different colors, but it does not directly indicate the difference in color. The measurement methods of CIELAB (the full name is CIE 1976 L*, a*, b*) are increasingly common and are increasingly used in the entire workflow from prepress to printing. CIELAB is used by software such as Photoshop and QuarkXPress. The color management technology through the ICC profile is also based entirely on CIELAB. CIELAB can now be used to measure colors on a variety of different media, not only on film, print, and print proofs, but also on printed proofs and LCD and CRT computer monitors. CIELAB is defined by a separate organization, the Commission International de l'Eclairage (CIE). As a CIE system, CIELAB simplifies the exchange of colors, because both parties can communicate

The term discusses colors instead of saying "prints are too blue" or "images are coloured." Since the CIE system is an international organization, these numbers used throughout the world represent the same color. CIELAB is the most common CIE color system.

The Color tool in Stemmer Imaging's CVB (Common Vision Blox) is a color search and color separation tool that utilizes the CIELAB color space.

The Color tool lets the user define a color template in the sample image, learn the colors, and then find the colors from the target image. The method of finding the color is to calculate the color difference between the color average of a given fascinating area or the color value of a single pixel (according to the selection of the search mode) and the color template of the training. The color template closest to the given area is the best matching template. In addition to measuring color differences, we also need to detect deviations under different lighting conditions.

Create a color training set (TrainingSet) with Color Teach software. The color TrainingSet is a collection of color samples used to describe user-defined colors. This can be done with the defined area in the sample image. Each area contains a color or what the human eye looks like is a color. This area becomes an instance. One or more instances of a color make up a variety of color templates.

Learning Classifier: The color of the instance represents the color space converted from RGB color space to CIELAB. From all pixels of a template, calculate the color center, the best color and deviation. The maximum deviation of a template pixel from the center of the color is denoted by MaxDistance, and the minimum deviation is denoted by MinDistance. The three parameters are used to calculate the quality of the template. A filter can be used to identify each pixel in the area. Use the filter to scan the original area. Calculate the quality of the pixel color template. Then process the image in filter mode.

The key to applying this tool is how to create a suitable template and set the maximum distance between different colors.

Note: You cannot use a solid color as a template because the minimum distance is equal to the maximum distance.

application:
(1) Fabric color quality inspection: mixed with other color impurities on solid fabrics, apply background color as filter,
Under the effect of the filter, the background color of the output image remains unchanged. The other color impurities are black, and the image is sequentially subjected to grayscale processing and binarization processing so that the impurities are completely separated from the background. Then use the blob tool to determine if the impurities meet the conditions.
(2) Recognition of cheques: The identification of cheques is mainly based on the identification of red seals, the use of color filters, the extraction of seals, or the positioning of seals, and the use of manto to identify the specific information of seals.
(3) Aerial image analysis: Using the different colors, the greening rate, the maturity of crops, or the area of ​​insects in forest trees can be calculated.

Source: China Vision Network