Colorization matrix construction with high compression efficiency for colorization-based coding using optimization

Image colourisation by non‐local total variation method in the CB and YIQ colour spaces

“…Let C denotes the colour difference channel I or Q, C 0 is the given colour information in the I or Q channel and λ > 0 is a given parameter. We propose the second model for image colourisation 16) where the parameter λ^ is defined by (3) and the non-local weight ω is defined as in (4) by replacing B with Y. Remark that the model (16) is different from the model (2) in the two aspects.…”

“…We propose the second model for image colourisation 16) where the parameter λ^ is defined by (3) and the non-local weight ω is defined as in (4) by replacing B with Y. Remark that the model (16) is different from the model (2) in the two aspects. Firstly, the model (16) is componentwise since I and Q are processed separately.…”

“…By introducing an extra variable C, we approximate the problem (16) by the following problem: 17) where θ is big enough to ensure that C is close to C. Then C and C can be solved by the alternating minimisation method. For fixed C, it is easy to get the solution of C:…”

“…Lee et al [15] formulate the colourisation-based compression problem into an optimisation problem by constructing the colourisation matrix in a multiscale manner. Mishiba and Yoshitome [16] propose a colourisation method by constructing a matrix with high compression efficiency and using optimisation for the colourisation-based coding. Pang et al [17] propose a hybrid scheme that combines both the example-based colourisation and the scribble-based colourisation algorithms.…”