Preprocessing
In general, images must be preprocessed before patterns can be recognized or objects can be found. In addition to calibration and noise reduction, projections and transformations such as shifting and rotating to the correct position might also be needed.
Image Analysis Algorithms
Depending on the task, algorithms for edge detection, segmentation, pattern recognition and feature extraction are used This can include a multiscale analysis (wavelets) or mathematical morphology.
Classification
For the classification of images, Convolutional Neural Networks (CNN) can deliver good results. If the images are parameterized, Boosted Decision Trees (BDT), Support Vector Machines (SVM) or other machine learning algorithms can be used.
Individual Solutions
Although there are extensive libraries and programs for this, the tasks are often too special and an individual solution is needed, For example, if the pixels are not square, the color depth is not 8 bit, or if the images are multidimensional (not just grayscale or RGB). Also, if high speed and a low memory footprint of the programs are required, it can be worthwile tailoring an algorithm specially to the problem.
Contact
For help with image processing, 256.systems can be contacted.