Gå til hovedinnhold
Publisert 2006

Read in English

Publikasjonsdetaljer

Tidsskrift : Food Hydrocolloids , vol. 20 , p. 650–662 , 2006

Internasjonale standardnummer :
Trykt : 0268-005X
Elektronisk : 1873-7137

Publikasjonstype : Vitenskapelig artikkel

Bidragsytere : Christiansen, K. F.; Krekling, Trygve; Kohler, Achim; Vegarud, Gerd Elisabeth; Langsrud, Thor; Egelandsdal, Bjørg

Har du spørsmål om noe vedrørende publikasjonen, kan du kontakte Nofimas bibliotekleder.

Kjetil Aune
Bibliotekleder
kjetil.aune@nofima.no

Sammendrag

Dressings were produced according to a fractional factorial design with protein type, protein level, oil level, pH, addition of NaCl, CaCl2 and sucrose, and processing temperature as design variables. The dressings were produced by high-pressure, and were model-systems emulsified and stabilized with different whey protein types.

The dressings were characterized by sensory analysis with regard to smell and taste, and texture properties. Analysis of variance revealed that oil level explained 56% of the variation in the sensory attribute viscosity. Protein type explained the largest part of the sensory attributes syneresis and smell. Addition of CaCl2 increased syneresis and gave a bitter taste.

The dressings' microstructures were captured by Scanning electron micrographs (SEM). The images were analysed using two different algorithms for feature extraction; the ABDF method, measuring absolute differences in greyness between pixels at fixed distances, and a local box-counting (fractal) method measuring the maximum differences in greyness within boxes of fixed sizes. After vectorization of the images by the algorithms, the vectors were analyzed to find which design variables influenced the vectors the most.

The analysis of variance revealed that protein type was the design variable that could clarify the largest part of the variation that could be explained in the images, followed by addition of NaCl. The fat was not visible in the images, and was hardly recognized by the feature extraction algorithms.

The correlation between viscosity and image analysis was fair, as the oil was not detected by the images or vectorized images. Sensory attributes explained by protein type or other design variables visible in the images were well explained by images. (c) 2005 Elsevier Ltd. All rights reserved.