Manifold-preserving image colorization with nonlocal estimation

Image colourisation using linear neighbourhood propagation and weighted smoothing

“…The proposed algorithm was compared with the methods in [2–4, 8]. Considering the requirements of both qualitative and quantitative indicators, the same stroke was applied to compare the different results for the same image, as shown in Fig.…”

“…5 d . In [8], manifold‐preserving was used for the colourisation, it sometimes have the colour‐crossing problems, as shown in the red box in Fig. 5 e .…”

“… Comparison of colourisation results of the present algorithm with those of three other methods (a) Artificially coloured image, (b)–(f) Colourisation using method in [2–4, 8] and the present algorithm…”

“…To illustrate the effect of the colour smoothing stage employed in this study, the obtained results were compared with those of the methods in [2–4, 8]. As can be observed from Fig.…”

“…Recently, some scholars have proposed a new image editing method. In [8], a novel stroke‐based manifold‐preserving image colourisation based on non‐local estimation is presented, which employs non‐local nearest neighbours to improve the efficiency. In [9], Deshpande et al exploited a LEARCH framework to train a quadratic objective function in the chromaticity maps, images are then colourised by minimising this objective function.…”