Automatic multi-resolution spatio-frequency mottle metric (sfmm) for evaluation of macrouniformity

Evaluation of mottle is an area of on-going research in print quality assessment. We propose an unsupervised evaluation technique and a metric that measures mottle in a hard-copy laser print. The proposed algorithm uses a scanned image to quantify the low frequency variation or mottle in what is supposed to be a uniform field. `Banding\u27 and `Streaking\u27 effects are explicitly ignored and the proposed algorithm scales the test targets from Flat print (Good) to Noisy print (Bad) based on mottle only. The evaluation procedure is modeled as feature computation in different combinations of spatial, frequency and wavelet domains. The model is primarily independent of the nature of the input test target, i.e. whether it is chromatic or achromatic. The algorithm adapts accordingly and provides a mottle metric for any test target. The evaluation process is done using three major modules: (1) Pre-processing Stage, which includes acquisition of the test target and preparing it for processing; (2) Spatio-frequency Parameter Estimation where different features characterizing mottle are calculated in spatial and frequency domains; (3) Invalid Feature Removal Stage, where the invalid or insignificant features (in context to mottle) are eliminated and the dataset is ranked relatively. The algorithm was demonstrated successfully on a collection of 60 K-Only printed images spread over 2 datasets printed on 3 different faulty printers and 4 different media Also, it was tested on 5 color targets for the color version of the algorithm printed using 2 different printers and 5 different media, provided by Hewlett Packard Company