Intelligent segmentation of jacquard warp-knitted fabric using a multiresolution Markov random field with adaptive weighting in the wavelet domain

In this paper, we propose a new approach to intelligently segment jacquard warp-knitted fabric images by combining wavelet texture decomposition, multiresolution Markov random field (MRF) modeling and Bayesian parameter estimation. Firstly, we use a flat scanner to capture eight-bit grayscale images of jacquard fabrics and adopt the Gaussian low-pass filter to decrease pixel variation due to discordant light reflection arising from uneven jacquard fabric surface. To overcome the incapacity of single resolution, multiresolution wavelet texture decomposition is employed, inspired by human visual sense procedure. Next, both intra-scale and inter-scale information are taken into account by the MRF model, in which a modified feature field model of the MRF, containing spatial noise with zero-mean Gaussian distribution, is presented in light of the inherent characteristics of jacquard warp-knitted fabric image. Afterward, an adaptive weighting function is used to weaken the defect of the potential parameter set empirically during the process of parameter estimation and image segmentation. Experimental results, used to verify the performance of the proposed algorithm especially the main novelties, prove that the approach is feasible and applicable.