The first stage of any vision related system is the image acquisition stage. In image processing, it is defined as the action of retrieving image from some source by using image sensor. It is the very first step of the whole process, and there is a need to capture the image with some satisfactory conditions. The images should be captured with a good quality and proper lightning conditions at the very first stage to make the further processing possible. However, if the images are not acquired satisfactorily then the intended task may not be achievable, even with the aid of image enhancement techniques. Therefore, in image processing it is a good practice to capture good quality images. Once the images are acquired then image-processing steps are implemented systematically.
The work presented in this chapter describes the processing steps systematically for diameter and length calculations as well as for the shape analysis.
4.3. Image Acquisition Setup
Setting up the imaging environment is a critical step in any imaging system. If the setup is built properly then we can save precious preprocessing time. Matlab also provides an image acquisition toolbox, which allows the real time processing, but in our application, it was not possible to install image-processing system in the spraying application center. Therefore, an offline high-speed camera imaging is installed in the application center. The following figure shows the setup of image acquisition system at BASF. Images are captured at the frame rate of 20,000 images per second.
At the lowest level of abstraction, when some operations on images are used such that, both input and output are intensity image. This is usually known as pre-processing of the images. Pre-processing does not increase image information content. If the information is measured using entropy, then pre-processing usually decreases it. Therefore, the best pre-processing is no pre-processing, and the way to achieve this is to concentrate on high quality image acquisition. Nevertheless, pre-processing is useful in a variety of ways since it can either suppress information irrelevant to our application or enhances some features important for further processing.
There are plenty of functions and operations for image pre-processing; however, we will discuss here only those, which fulfilled the purpose of our application.
As we can see that, if image is segmented without pre-processing, some part of the image was detected and some part was not. Light is an important factor in our application, which hinders the detection of all features. Therefore, a good pre-processing technique is devised here, which allows the full feature extraction in further processing.
First, we need to understand the nature of images to select the suitable operations. As the spraying images are usually filaments (lines) and there is quick transition of color in filaments and background of the image. Therefore, the contrast can be enhanced using sharpening function. Sharpness is actually the contrast between different colors. A quick transition from black to white looks sharp. A gradual transition from black to gray to white looks blurry. Sharpening images increases the contrast along the edges where different colors meet. Therefore, in the first step image contrast is enhanced using “imsharpen” function in Matlab.
Secondly, the noise from the image is removed by using Weiner filter. Wiener estimates the local mean and variance around each pixel. It filters the grayscale image using a pixel-wise adaptive low-pass Wiener filter. The Wiener filter tailors itself to the local image variance. Where the variance is large, it performs little smoothing. Where the variance is small, it performs more smoothing. This approach often produces better results than linear filtering. The adaptive filter is more selective than a comparable linear filter, preserving edges and other high-frequency parts of an image. However, Wiener is computationally expensive than linear filters.
Finally, Histogram adjustment is a very useful technique to enhance image. Contrast Limited Adaptive histogram is nothing but an alternative of normal histogram equalization, which performs the action on entire image while adaptive histogram works on small parts of the image known as tiles. The contrast of each tile is enhanced separately and then tiles are combined again to reduce any artifacts that may appear in the image. The contrast, especially in homogeneous areas, can be limited to avoid amplifying any noise that might be present in the image. The whole variation of images is shown in the figure. By using these techniques for pre-processing, we were able to detect full features of the images. The next step is to segment the images is discussed in the section 4.4.