Image Processing

Image pre-processing

Today the fields of application for image processing are many and varied. With the growing high requirements for automated systems, quality and quantity as well as processing speed play a large role. Thus an optimisation of the procedures of image processing is necessary in various respects.

Image pre-processing serves to prepare the input image data in order to generate the most optimal image data possible for subsequent image processing (e.g. object recognition). When processing the raw data of an image sensor, the following steps are required in order to obtain a typical colour picture:

• Demosaicing (colour reconstruction of the Bayer colour mask)
• Contrast and gamma correction
• White balancing, colour correction

Subsequently, other image pre-processing steps can be followed which are associated with image restoration and image enhancement:

• Reduction of noise or disturbances
• Sharpening or smoothing of edges
• Correction of contortions of perspective

Object recognition

Object recognition constitutes an umbrella term for the identification of objects in images or video sequences. The complete process of object recognition can be subdivided into segmentation, extraction of features and classification.

High Dynamic Range

When speaking of High Dynamic Range Imaging (HDRI), one refers to the generation of images with an increased dynamic range (HDR). Within a motif, different areas are variously light or respectively dark. Whether we are dealing with high or low dynamic range is determined by the size of this difference. Conventional cameras are not capable of reproducing the differences in brightness which occur in nature in true detail, and therefore tend towards under or overexposure. A broader dynamic range can first be achieved through the usage of special HDR image sensors or by overlaying multiple individually-taken images with different lengths of exposure.