Image Analysis & Processing at Avantier

Most images nowadays are recorded in digital form. From a mathematical point of view, they are just 2-dimensional or multidimensional arrays containing values of pixel signals. In simplest binary images each pixel can have only 2 values: zero and one. Gray-scale (black and white, monochrome) or indexed images have one value per pixel, and the value usually is 8-bit or 16-bit, while any other number of bits also may be used. In addition to the image array, color indexed images also have a color table, which assigns at least 3 (Red, Green, and Blue) color values to each image array value. More common RGB color images just contain Red, Green, and Blue intensity values (typically 8-bit ones) for each pixel. Multispectral/Hyperspectral images may contain any number of intensity values per pixel obtained in different spectral ranges. These values often have 16 bits, but may have any number of bits, and may be recorded as float or double values. Uncompressed image arrays usually are saved in BMP or TIFF or RAW files. Industrial and scientific images normally are not compressed so they preserve all the image details. Consumer quality images commonly are recorded as highly compressed JPEG files, which discard a lot of almost non-noticeable information contained in RAW images. Medical imaging uses specialized file formats, such as DICOM. Say, MRI or CT scan results are saved in DICOM files as a sequence of many gray-scale 2-dimensional arrays. Computer generated images, unlike natural images, do not have noise, and so are very suitable for lossless compression in PNG files.