Locally Adaptive Processing of Color Tensor Images Represented as Vector Fields

A new approach for locally-adaptive processing of color RGB images represented as tensors of size M × N × 3, is offered in this work. Unlike the famous similar methods of the kind, the processing here is executed on a single matrix only, which comprises the modules of the vectors, corresponding to the image pixels’ colors. A group of related basic algorithms for locally-adaptive processing is presented, which have lower computational complexity than that of the algorithms, applied individually on each of the RGB components. As it is known, in the famous color RGB transform models of the kind YCrCb, HSV, HSI, Lab, KLT, etc., the processing is applied on the most powerful transformed color component only, and after inverse operation, the original RGB model is restored. In contrast, the idea for locally-adaptive processing does not need direct and inverse transform of the color model. Together with this, the brightness and the color hue of the processed image pixels, are retained. The characteristics of the proposed basic algorithms for contrast enhancement, linear and non-linear sharpness filtration, noise suppression, and texture segmentation, are defined. Some examples for locally-adaptive processing of color medical images are given, which illustrate the related algorithms. The presented approach could be also used for other kinds of color images, where the visibility of their local structure is of high importance.