Automated Linear-Time Detection and Quality Assessment of Superpixels in Uncalibrated True- or False-Color RGB Images

— In this methodological paper, provided with a relevant survey value, an original low-level computer vision (CV) software pipeline, called RGB Image Automatic Mapper (RGBIAM), is presented and discussed. RGBIAM is a lightweight computer program capable of automated near real-time superpixel detection and quality assessment in an uncalibrated monitor-typical red-green-blue (RGB) image, depicted in either true-or false-colors. The RGBIAM system design consists of known CV modules, constrained by the Calibration/Validation (Cal/Val) requirements of the Quality Assurance Framework for Earth Observation (QA4EO) guidelines. In agreement with the QA4EO Cal requirements, to benefit from multi-source data harmonization and interoperability, RGBIAM requires as mandatory an uncalibrated RGB image pre-processing first stage, consisting of an automated (self-organizing) statistical model-based color constancy algorithm. The RGBIAM's hybrid (combined deductive/top-down and inductive/bottom-up) inference pipeline comprises: (I) a direct quantitative-to-nominal (QN) RGB variable transform, where RGB pixel values are mapped (quantized) onto a prior dictionary of color names, to be community-agreed upon in advance, equivalent to a static (non-adaptive to data) polyhedralization of the RGB cube. Prior color naming is the deductive counterpart of popular inductive vector quantization (VQ) algorithms, whose typical VQ error function to minimize is a root mean square error (RMSE). In the output multi-level color map domain, superpixels are automatically detected in linear time as connected sets of pixels featuring the same color label. (II) An inverse nominal-to-quantitative (NQ) RGB variable transform, where a superpixelwise-constant RGB image approximation is generated in linear time, which allows to assess a VQ error image in compliance with the QA4EO Val requirements. The hybrid direct and inverse RGBIAM-QNQ transform is: (i) general-purpose, i.e., data-and application-independent. (ii) Automated, i.e., it requires no user-machine interaction. In the hybrid RGBIAM pipeline, a deductive inference first stage, analogous to genotype, provides automatically inherently ill-posed inductive learning-from-data algorithms, equivalent to phenotype, with initial conditions. (iii) Near real-time, i.e., its computational complexity increases linearly with the image size. (iv) Implemented in tile streaming mode, to cope with massive images. As a proof-of-concept, a realization of the RGBIAM pipeline was tested on three RGB images acquired by different imaging sensors and acquisition platforms. Collected outcome and process quantitative quality indicators, including degree of automation, computational efficiency, VQ rate and VQ error, are consistent with theoretical expectations and reveal that the RGBIAM lightweight computer program is suitable for low-level CV mobile software applications specifically designed to run on web browsers and mobile devices, such as tablet …